We aimed to establish the prevalence and effect of worsening renal function (WRF) on survival among patients with acute decompensated heart failure. Furthermore, we sought to establish a risk score for the prediction of WRF and externally validate the previously established Forman risk score. A total of 657 consecutive patients with acute decompensated heart failure presenting to the emergency department and undergoing serial creatinine measurements were enrolled. The potential of the clinical parameters at admission to predict WRF was assessed as the primary end point. The secondary end point was all-cause mortality at 360 days. Of the 657 patients, 136 (21%) developed WRF, and 220 patients had died during the first year. WRF was more common in the nonsurvivors (30% vs 41%, p = 0.03). Multivariate regression analysis found WRF to independently predict mortality (hazard ratio 1.92, p <0.01). In a single parameter model, previously diagnosed chronic kidney disease was the only independent predictor of WRF and achieved an area under the receiver operating characteristic curve of 0.60. After the inclusion of the blood gas analysis parameters into the model history of chronic kidney disease (hazard ratio 2.13, p = 0.03), outpatient diuretics (hazard ratio 5.75, p <0.01), and bicarbonate (hazard ratio 0.91, p <0.01) were all predictive of WRF. A risk score was developed using these predictors. On receiver operating characteristic curve analysis, the Forman and Basel prediction rules achieved an area under the curve of 0.65 and 0.71, respectively. In conclusion, WRF was common in patients with acute decompensated heart failure and was linked to significantly worse outcomes. However, the clinical parameters failed to adequately predict its occurrence, making a tailored therapy approach impossible.
Worsening renal function (WRF) is a common complication during the treatment of acute decompensated heart failure (ADHF) and has even earned the term “cardiorenal syndrome type I.” About 20% to 40% of patients hospitalized for ADHF concurrently develop WRF. The increased short-term morbidity, treatment costs, and in-hospital mortality of patients with heart failure who develop WRF have been well established. Recent studies have also showed WRF to possess a graded association with long-term mortality. Although WRF has been reported to occur more frequently in the presence of co-morbid conditions, such as impaired baseline renal function, lowered left ventricular function, age, and arterial hypertension, and a clinical risk score has been advocated, its prediction remains a daily clinical challenge. We, therefore, aimed to determine the clinical predictors of WRF in a large cohort of patients with ADHF presenting to the emergency department. Furthermore, we sought to establish a novel risk score for the prediction of WRF and externally validate the previously established Forman risk score.
Methods
During the enrollment periods (from May 2001 to April 2002 and April 2006 to April 2010), we prospectively screened all patients presenting to the participating emergency departments (University Hospital, Basel, Kantonsspital Aarau, Kantonsspital, Lucerne) with a chief complaint of acute dyspnea. A total of 767 patients were diagnosed with ADHF according to the current guidelines. The 657 patients with ADHF hospitalized for >48 hours and undergoing serial creatinine measurements constituted the study population for the present analysis. The 1-year follow-up data were complete for 575 patients. The patients <18 years of age, those undergoing hemodialysis, and trauma cases were excluded. The study was performed according to the principles of the Declaration of Helsinki. The local ethical committee approved the present study. All participants provided written informed consent. At enrollment, all patients underwent an initial clinical assessment, including clinical history, physical examination, pulse oximetry, blood tests, and chest radiography. Echocardiography, blood gas analyses, and pulmonary function tests were performed according to the treating physicians’ recommendations.
In accordance with the clinical practice guidelines for chronic kidney disease (CKD) from the American National Kidney Foundation and for comparison with previous studies, renal function was divided into 3 groups according to the glomerular filtration rate (GFR; <30 ml/min/1.73 m 2 , 30 to 60 ml/min/1.73 m 2 , and >60 ml/min/1.73 m 2 ) on presentation to the emergency room. The GFR was calculated using the abbreviated Modification of Diet in Renal Disease Study (MDRD) equation. This method is currently considered the best method clinically available for this purpose and has been validated in patients with advanced HF. WRF was defined as an in-hospital increase in serum creatinine >0.3 mg/dl (26.5 μmol/L) from admission. We defined CKD as a stably reduced estimated GFR of <60 ml/min/1.73 m 2 that had persisted for >3 months before hospitalization.
The establishment of a tool to stratify patients with ADHF at admission according to their risk of developing WRF during the hospitalization, as well as the potential of this score and the Forman risk score to predict WRF, was the primary end point of the present study. All-cause mortality at 360 days of follow-up was assessed as the secondary end point. Mortality was prospectively assessed during follow-up. Patients were interviewed by telephone at 6 and 12 months after their initial presentation. In addition, referring physicians and the administrative databases of the respective hometowns were interviewed/accessed in the case of uncertainty regarding the health status or additional hospitalizations.
The ADHF diagnosis was adjudicated by 2 independent cardiologists not involved in the emergency department care using all available medical records pertaining to the individual patient, including the response to therapy and the autopsy data for those patients who had died in hospital.
The statistical analyses were performed using Statistical Package for Social Sciences for Windows, version 15.0 (SPSS, Chicago, Illinois). A statistical significance level of p <0.05 was used. The comparison between the 2 groups was done using the chi-square test and Fisher’s exact test for categorical variables and the t test for continuous variables if normally distributed or the Mann-Whitney U test if not normally distributed. Cumulative survival was calculated using Kaplan-Meier analysis and differences between the curves were evaluated using log-rank statistics.
The independent predictors of WRF were identified using multivariate regression analysis. The analyzed variables included age, gender, vital signs, co-morbidities, outpatient medications, and laboratory parameters. The laboratory parameters included the left ventricular ejection fraction, blood gas analysis parameters, and cardiac troponin T and B-type natriuretic peptide levels. Overall, 48 parameters were evaluated. The variables were entered at an entry level of significance p <0.2 and kept in the model at an existence level of p <0.05. A clinically applicable risk score was calculated as the sum of point values assigned to each independent predictor variable. The number of points assigned to each predictor depended on the regression coefficient β. Because the regression coefficient β values of all parameters in the final model were similar, no weighing of the point values (i.e., more points for certain parameters) was necessary. For continuous variables, a point value was assigned for values above (1) or below (0) the best calculated cut-off value using Youden’s J index. In the complex, computer-based risk score model, continuous values and actual regression coefficient values were used. In a first step, the baseline characteristics, vital signs, and standard laboratory parameters available for all 522 patients were assessed. In a second step, the blood gas analysis parameters were added to the analysis; these parameters were available for 223 patients. Furthermore, the previously described Forman prediction rule was calculated. The relation among the Basel risk score, Forman risk score, and development of WRF was assessed using the Jonckheere-Terpstra test. The prognostic accuracy of the different models was evaluated using receiver operating characteristic (ROC) curve analysis. The areas under the ROC curves were compared using MedCalc software, version 9.2 (MedCalc Software, Mariakerke, Belgium).
Results
The detailed baseline characteristics of the study population are listed in Table 1 . Overall, 136 patients (21%) experienced WRF during the hospitalization. No differences were seen in the co-morbidities, vital signs, outpatient treatment, admission laboratory parameters, or the incidence of WRF between the patients undergoing blood gas analyses and those without blood gas analyses.
| Variable | All Patients (n = 657) | WRF | p Value | |
|---|---|---|---|---|
| No (n = 521) | Yes (n = 136) | |||
| Age (years) | 79 [71–85] | 79 [70–85] | 79 [72–85] | 0.36 |
| Women | 293 (45%) | 240 (46%) | 53 (39%) | 0.08 |
| Arterial hypertension | 460 (70%) | 364 (70%) | 96 (71%) | 0.55 |
| Heart failure | 356 (55%) | 274 (54%) | 82 (61%) | 0.59 |
| Coronary artery disease | 353 (54%) | 281 (54%) | 72 (53%) | 0.49 |
| Diabetes mellitus | 196 (31%) | 148 (29%) | 48 (36%) | 0.08 |
| Chronic kidney disease | 268 (42%) | 190 (37%) | 78 (58%) | <0.01 |
| Neoplastic disease | 137 (21%) | 111 (21%) | 26 (19%) | 0.89 |
| Diuretics at admission | 456 (70%) | 349 (67%) | 107 (79%) | 0.01 |
| β Blockers at admission | 359 (56%) | 275 (54%) | 84 (63%) | 0.04 |
| Nitrates at admission | 157 (24%) | 127 (24%) | 30 (22%) | 0.95 |
| Renin angiotensin system blockers at admission | 390 (60%) | 298 (58%) | 82 (61%) | 0.40 |
| Digitalis at admission | 63 (10%) | 53 (10%) | 10 (11%) | 0.33 |
| Aspirin at admission | 305 (47%) | 248 (47%) | 57 (42%) | 0.76 |
| Calcium channel blockers at admission | 146 (21%) | 100 (20%) | 33 (24%) | 0.50 |
| Anticoagulation at admission | 253 (39%) | 198 (38%) | 55 (41%) | 0.34 |
| Blood pressure (mm Hg) | ||||
| Systolic | 137 [120–160] | 136 [120–158] | 144 [125–164] | 0.02 |
| Diastolic | 83 [70–96] | 83 [70–96] | 84 [74–97] | 0.54 |
| Heart rate (beats/min) | 91 [75–111] | 96 [77–111] | 89 [71–120] | 0.73 |
| Oxygen saturation (%) | 96 [92–98] | 95 [90–98] | 93 [90–98] | 0.86 |
| Left ventricular ejection fraction (%) | 40 [30–60] | 45 [30–60] | 40 [30–55] | 0.69 |
| Urea (mmol/L) | 11.0 [6.9–18.5] | 10.0 [6.6–18.0] | 12.6 [8.6–22.1] | <0.01 |
| Creatinine (μmol/L) | 107 [80–148] | 104 [78–146] | 115 [86–156] | 0.02 |
| Glomerular filtration rate (ml/min) | 49 [32–71] | 50 [33–72] | 49 [21–68] | 0.02 |
| B-type natriuretic peptide (pg/ml) | 981 (534–1493) | 925 [520–1441] | 1,071 (709–1714) | 0.02 |
| Troponin T | 0.04 [0.01–0.4] | 0.03 [0.01–0.355] | 0.03 [0.01–0.300] | 0.14 |
| pH ⁎ | 7.42 [7.35–7.45] | 7.42 [7.36–7.46] | 7.41 [7.35–7.44] | 0.06 |
| Bicarbonate (mmol/L) ⁎ | 23.2 [20.6–26.9] | 24.0 [21.0–24.5] | 21.7 [18.2–25.4] | 0.84 |
| Percutaneous coronary intervention during hospitalization | 70 (11%) | 55 (11%) | 15 (11%) | 0.87 |
| Contrast-enhanced computed tomography during hospitalization | 103 (16%) | 80 (15%) | 23 (17%) | 0.98 |
| Hospital stay (days) | 13 [7–19] | 14 [9–20] | 15 [8–23] | 0.42 |
| In-hospital mortality | 56 (9%) | 33 (6%) | 23 (17%) | <0.01 |
| 360-Day mortality ⁎ | 220 (38%) | 171 (30%) | 49 (41%) | 0.03 |
⁎ pH and bicarbonate levels available for 256 patients and 360-day mortality for 575 patients.
The risk factors for the development of WRF are listed in Table 2 . Previously diagnosed CKD was the only independent predictor of WRF. The creatinine values (p = 0.38), urea values (p = 0.87), GFR rate (p = 0.74), or GFR classes (p = 0.98) did not independently predict WRF. In a ROC curve analysis for the development of WRF, previously diagnosed CKD achieved an area under the curve of 0.60 (95% confidence interval [CI] 0.55 to 0.66). However, only 28% of patients with previously diagnosed CKD developed WRF compared to 15% of patients without known CKD. Furthermore, 42% of all WRF cases developed in patients without known CKD.
| Predictor | Hazard Ratio (95% CI) | p Value |
|---|---|---|
| Univariate analysis | ||
| Glomerular filtration rate class | 1.35 (1.04–1.73) | 0.02 |
| History of diabetes | 1.36 (0.91–2.04) | 0.13 |
| History of chronic kidney disease | 2.31 (1.57–3.40) | <0.01 |
| Outpatient diuretics | 1.89 (1.15–2.82) | 0.01 |
| Outpatient β blockers | 1.44 (0.97–2.13) | 0.06 |
| Bicarbonate | 0.92 (0.87–0.97) | 0.03 |
| Multivariate analysis | ||
| History of chronic kidney disease | 2.07 (1.29–3.30) | <0.01 |
| Multivariate analysis including blood gas analyses (n = 253) | ||
| History of chronic kidney disease | 2.13 (1.08–4.19) | 0.03 |
| Outpatient diuretics | 5.75 (2.10–15.73) | <0.01 |
| Bicarbonate | 0.91 (0.86–0.97) | <0.01 |
Including the blood gas analysis parameters into the calculation, previously diagnosed CKD, outpatient treatment with diuretics, and lowered bicarbonate levels were independently associated with the development of WRF. Again, the creatinine levels (p = 0.60), urea values (p = 0.99), GFR rate (p = 0.52), and GFR classes (p = 0.32) failed to predict WRF. A risk score was developed using the predictors. One point each was assigned for outpatient diuretic treatment, known CKD, and bicarbonate level <21 mmol/L. The relation between the calculated Basel risk score and the development of WRF is listed in Table 3 . Of the 20% of patients in risk group 0, 2% developed WRF, and of the 43% of patients in risk groups 2 and 3, 29% developed WRF. However, >1/3 (37%) of all WRF cases developed in the low-risk (0) and intermediate-risk (1) groups. In a ROC curve analysis for the development of WRF, the Basel risk score achieved an area under the curve of 0.71 (95% CI 0.63 to 0.79), significantly surpassing the diagnostic accuracy of the single-parameter model (p = 0.02). A computer-based, complex, exponential Basel risk assessment model achieved an area under the curve of 0.75 (95% CI 0.67 to 0.82).
| Class | WRF | |
|---|---|---|
| No (n = 204) | Yes (n = 51) | |
| 0 | 49 (24) | 1 (1) |
| 1 | 78 (38) | 18 (35) |
| 2 | 64 (31) | 14 (27) |
| 3 | 13 (6) | 18 (35) |
The relation between the Forman risk score and the development of WRF is listed in Table 4 . Greater Forman risk scores were not associated with a greater incidence of WRF (0.09). Hence, 14% of patients in risk group 0 developed WRF compared to 27% of patients in risk groups 4 or greater. Additionally, >50% of WRF cases developed in the low-risk groups, risk groups 2 or less. In a ROC curve analysis for the prediction of WRF, the Forman risk score achieved an area under the curve of 0.65 (95% CI 0.57 to 0.71). The prognostic accuracy of the novel Basel score and Forman risk score was similar (p = 0.24). Overall, 56 patients died in hospital and 220 patients died during the 360 days after admission. Independent of the time of death, WRF was more common in patients who died during follow-up (in-hospital, p <0.01; 360 days, p = 0.03). Thus, multivariate Cox regression analysis found WRF to independently predict short-term ( Table 5 ) and long-term ( Table 6 ) mortality. Spot measurements of renal function described by the creatinine values, urea values, or continuous GFR values at admission failed to predict in-hospital mortality. Nevertheless, morbidity was greater in the patients with a lower GFR class, as reflected by the longer duration of the initial hospitalization (GFR <30 ml/min, 16 days, range 9 to 24; vs GFR 30 to 60 ml/min, 12 days, range 6 to 19; vs GFR >60 ml/min, 11 days, range 7 to 17; p <0.01). No relevant differences were seen in the discharge medication prescribed for HF among the patients with and without renal dysfunction. However, the 90- and 360-day mortality rates were significantly greater in patients with a lower admission GFR class (90-day, 28% vs 15% vs 12%, p <0.01; 360-day, 48% vs 33% vs 24%, p <0.01). As shown in Figure 1 , the risk of death clearly increased with decreasing GFR class. However, the occurrence of WRF further increased the risk of death independently of the baseline GFR class.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
