| | Worsening renal function: What is a clinically meaningful change in creatinine during hospitalization with heart failure?☆☆☆Received 26 July 2002; received in revised form 26 November 2002 and 27 November 2002 Abstract Introduction: Worsening renal function during hospitalization for heart failure, defined as elevation in creatinine during admission, predicts adverse outcomes. Prior studies define worsening renal function using various creatinine elevations, but the relative value of definitions is unknown. Methods and Results: In a prospective cohort of 412 patients hospitalized for heart failure, we compared a spectrum of worsening renal function definitions (absolute creatinine elevations ≥0.1 to ≥0.5 mg/dL and 25% relative elevation from baseline) and associations with 6-month mortality, readmission, and functional decline. Creatinine elevation ≥0.1 mg/dL occurred in 75% of patients, and elevation ≥0.5 mg/dL occurred in 24% of patients. Risk of death rose with higher creatinine elevations (adjusted hazard ratio [HR] = 0.89, 1.19, 1.67, 1.91, and 2.90 for elevations ≥0.1 to ≥0.5 mg/dL). Maximum sensitivity of any definition for predicting mortality was 75% and maximum specificity was 79%. High creatinine elevation was a more important predictor of death than was a single measure of baseline creatinine. Conclusions: Larger creatinine elevations predict highest risk of death, yet even minor changes in renal function are associated with adverse outcomes. The choice of a “best definition” for worsening renal function has implications for the number of patients identified with this risk factor and the magnitude of risk for mortality.
Creatinine levels often rise in patients hospitalized with heart failure (HF), and this change is increasingly recognized as having prognostic importance. In elderly HF patients, creatinine elevations of at least 0.3 mg/dL during admission have been associated with longer hospital stay, higher hospitalization costs, and increased in-hospital and long-term mortality.1 Previous studies focusing on HF patients, however, have not established whether even smaller increases in creatinine could be clinically significant, as indicated by independent associations with adverse outcomes after discharge. Studies of worsening renal function (WRF) in other patient populations use varying levels of creatinine elevation to define renal dysfunction and often focus on extreme increases, examining the effects of creatinine elevations of 0.5 mg/dL and higher on outcomes.2, 3, 4, 5, 6, 7, 8, 9, 10 In HF patients, even a slight elevation in creatinine could be important, but little evidence is available to judge the optimal definition of clinically significant WRF during hospital admission.
Identifying a clinically meaningful definition of WRF could have important practical implications for managing hospitalized HF patients. To address this issue, we determined how various definitions of WRF, defined as elevations in creatinine, were associated with mortality, readmission, and functional decline in the 6 months after discharge. We also established the sensitivity and specificity of these definitions and investigated whether impaired baseline renal function, measured by the creatinine level at discharge, modified the relationship between WRF and poor outcomes.
Methods  Study sample We screened consecutive patients admitted to Yale-New Haven Hospital between March 1996 and September 1998 who were ≥50 years of age and met clinical criteria for presence of HF on admission. To identify eligible patients, admissions were screened daily in 2 phases. First, we identified patients with either an admission diagnosis of HF or radiologic signs of HF on the admission chest x-ray. Second, patients who met either of the above conditions had their medical records reviewed within 3 days of admission to verify the presence of HF, based on a modification of the first National Health and Examination Survey (NHANES-I)11 and criteria by Harlan et al.12 Symptoms include dyspnea, exertional dyspnea, orthopnea, paroxysmal nocturnal dyspnea, cough/nocturnal cough, fatigue, and confusion; signs include S3, rales, hypotension, cardiogenic shock, cardiac arrest, respiratory rate ≥24, peripheral edema, tachyarrhythmias, or conduction disorders. Details have also been published elsewhere.13 We excluded patients who were admitted without evidence of HF, patients transferred from other hospitals or admitted from nursing homes (for consistency of baseline medical record abstraction data), patients with a noncardiovascular terminal illness such as cancer with <6-month expected survival (for follow-up purposes), and patients with HF secondary to high-output states or noncardiac diseases (for homogeneity of study population). Of the 1,151 patients who were screened, 648 met eligibility criteria. Patients eligible but not enrolled were previously enrolled or in other studies (126), died or were discharged before interview (45), refused or were unable to be interviewed (45), or lived in another state (6). We excluded 5 patients because of missing baseline interview or medical records. We also excluded 2 patients with missing creatinine levels and 7 patients who died in-hospital (2%), leaving 412 patients in the present analysis. Worsening renal function and covariates Serum creatinine levels on admission and at discharge, as well as peak creatinine during admission, were abstracted from patient medical records. We evaluated the following definitions for WRF: serum creatinine elevation during admission by ≥0.1, ≥0.2, ≥0.3, ≥0.4, and ≥0.5 mg/dL from first admission value. We also used another common definition of WRF: a 25% increase in creatinine to a peak value of at least 2.0 mg/dL.9, 10, 14 We also abstracted demographic variables, clinical history, admission characteristics, diagnostic testing, treatments, and discharge medications from the medical records. Left ventricular ejection fraction (LVEF) was determined from quantitative or qualitative assessments from radionuclide ventriculography, cardiac catheterization, or echocardiography performed during or within 1 year of the index hospitalization. In cases in which multiple assessments were available, priority was given to the most recent quantitative assessment. Outcomes Primary outcomes included time to death and time to first hospital readmission for any cause. Deaths were ascertained through next of kin, hospital records, and active monitoring of obituaries. An assessment of hospital readmissions was conducted using hospital administrative databases and discharge summaries to determine cause of readmission. A validation subsample of patients (21% of study sample) compared administrative readmission data with patient self-report for admissions at Yale-New Haven Hospital and other hospitals, which indicated that 95% of all readmissions occurred at Yale-New Haven Hospital. The other main outcome in our analysis was the combined endpoint of functional decline or death within 6 months of discharge from the index admission, and this variable was coded dichotomously. Functional status was assessed during follow-up interviews (6 months after discharge date), and of the surviving 341 patients at 6 months, more than 94% had complete functional status information. The proportion of patients without functional decline data was not significantly different for those with WRF compared with those without WRF (4% vs. 7%, P = .20). Functional decline was defined as the loss of at least 1 basic activity of daily living over the follow-up period. We used the Katz Activities of Daily Living15 scale to measure functional status at baseline and at follow-up and determined whether patients needed any help in performing the following basic activities: walking across a small room, bathing, dressing, eating, moving from bed to chair, and using a toilet. This instrument has high validity and reliability and has been shown to be appropriate for use in an acutely ill population.16 The combined outcome was used to avoid any bias resulting from potentially different rates of death in the 2 comparison groups. However, in addition, to isolate the effect, we conducted secondary analyses on the outcome of functional decline alone in the subset of survivors. Statistical analysis Unadjusted outcomes and sensitivity analysis Unadjusted associations for each of the cutpoints for WRF and outcomes were tested using the Pearson chi-square test and the Kaplan-Meier log-rank test. We calculated the sensitivity and specificity of all WRF definitions in predicting outcomes at 6 months after discharge. Mortality and readmission The effect of various definitions of WRF on adjusted risk ratios and odds ratios (ORs) was also analyzed. Multivariable Cox proportional hazards regression tested whether these definitions of WRF were important independent predictors of death, all-cause hospital readmission, HF hospital readmission, and the combined endpoint of death or readmission. Time to event was calculated from the date of discharge. Deaths were censored for models predicting readmission. Multivariable analyses included covariates identified by clinical judgment, bivariate analyses, and previous studies as important predictors of readmission17 and mortality.18, 19, 20 Variables considered included age, sex, race, cardiac history (hypertension, angina, myocardial infarction, HF), noncardiac history (stroke, diabetes), admission variables (systolic and diastolic blood pressure, LVEF), and medications at discharge (angiotensin-converting enzyme inhibitors, diuretics, β-blockers, calcium channel blockers, aspirin, digoxin). Final models in the Results section include only the most parsimonious models. In these models, age, LVEF, years of HF, and discharge creatinine were entered as continuous, independent variables. An interaction term for LVEF (categorized as <40% vs. ≥40%) was also tested for significance. Creatinine at discharge was considered the baseline value because the follow-up period began after discharge. Adverse events during admission, including cerebrovascular accident, shock, hypotension, cardiac arrest, myocardial infarction, pneumonia, and arrhythmia, were also added into models to compare the importance of WRF in predicting poor outcomes with these other in-hospital events. Proportionality assumptions were tested using a time-interaction term in the model, which was excluded from the final model if it was not significant. Predictors of functional decline or death Multivariable logistic regression analysis was used to determine the independent associations between the various definitions of WRF and risk of death or decline in activities of daily living after adjusting for covariates identified from bivariate analyses. Multivariate models were also constructed to examine whether WRF was an independent predictor of functional decline in the subset of patients who survived the 6-month follow-up period. Interaction terms and other analyses In all models, interaction terms for WRF and discharge creatinine, for WRF and first admission creatinine, and for WRF and peak creatinine were tested for significance. Discharge and first admission creatinine levels were categorized for these analyses using the following cutpoints: <1.0 mg/dL, 1.0 to <2.0, and ≥2.0. Peak creatinine, defined as the highest creatinine level reached during the admission, was categorized using the following cutpoints: <2.0 and ≥2.0. We tested independent effects of these variables as well as interaction effects with WRF, using the Breslow-Day test for homogeneity in bivariate analysis and interaction terms in multivariate analyses. We created stratified models for each stratum to confirm results derived from testing interaction terms. For all multivariable models, continuous covariates were tested for linearity, and those variables that did not exhibit a linear relationship with the outcome were recoded as categoric variables. For each model, diagnostic plots were generated to identify outliers and test model fit, and parsimonious models were identified. All analyses were conducted using SAS 6.12 (Statistical Analysis Software, Cary, NC). Results Study sample Patients in the study sample were generally elderly, with a mean age of 72 years (standard deviation [SD] 11). The sample was 51% male and 76% white. The majority of patients had a history of HF (72%) and other cardiovascular disease. Many patients had a history of diabetes (47%) and hospitalization for HF (38%). Mean LVEF was 39% (SD 17), with 47% of patients having an LVEF ≥40%. The mean value for creatinine on admission was 1.8 mg/dL (SD 1.4), and 75% of patients presented with a creatinine level <2.0 mg/dL (Fig. 1).
Only 7 patients (2%) were on dialysis at the time of admission. Worsening renal function The frequency of WRF for the cohort varied with the definition—from 24% for the most restrictive definition of creatinine elevation (≥0.5 mg/dL) to 75% for the most inclusive definition (≥0.1 mg/dL) (Table 1, Fig. 2).
For the definition of a relative increase by at least 25% from baseline admission level to 2.0 mg/dL or higher, the frequency of WRF was 10%. | | |  | WRF Definitions | Sample (%) | Died (%) | Sensitivity (%) | Relative* Sensitivity Change (%) | Specificity (%) | Relative* Specificity Change (%) | Unadjusted HR (95% CI) | Adjusted** HR (95% CI) | Adjusted† HR (95% CI) | |
|---|
| WRF (≥0.1) | 75 | 15 | 75 | — | 25 | — | 1.01 | 0.89 | 0.88 | |
| No WRF | 25 | 16 | | | | | (0.91, 1.11) | (0.50, 1.60) | (0.49, 1.57) | |
| WRF (≥0.2) | 58 | 17 | 64 | −15 | 43 | +74 | 1.25 | 1.19 | 1.15 | |
| No WRF | 42 | 13 | | | | | (0.78, 2.00) | (0.70, 2.04) | (0.67, 1.97) | |
| WRF (≥0.3) | 45 | 19 | 57 | −10 | 57 | +33 | 1.62 | 1.67 | 1.61 | |
| No WRF | 55 | 12 | | | | | (1.02, 2.57) | (0.98, 2.85) | (0.94, 2.77) | |
| WRF (≥0.4) | 32 | 21 | 44 | −23 | 70 | +23 | 1.68 | 1.91 | 1.83 | |
| No WRF | 68 | 13 | | | | | (1.07, 2.63) | (1.10, 3.31) | (1.05, 3.23) | |
| WRF (≥0.5) | 24 | 26 | 40 | −11 | 75 | +13 | 2.11 | 2.90 | 2.86 | |
| No WRF | 76 | 12 | | | | | (1.34, 3.31) | (1.61, 5.20) | (1.55, 5.26) | |
| WRF (≥25%) | 9 | 23 | 14 | −64 | 91 | +16 | 1.59 | 1.76 | 1.67 | |
| No WRF | 91 | 14 | | | | | (0.85, 2.98) | (0.83, 3.71) | (0.78, 3.56) | |
| *Relative change calculated as percent loss or gain compared with previous definition of WRF. **Adjusted for age, left ventricular ejection fraction, discharge creatinine, diabetes, and years of heart failure. †Adjusted for above clinical and demographic characteristics and in-hospital adverse events. | | | | |
Mortality Patients with WRF (defined as creatinine elevations between ≥0.2 mg/dL and ≥0.5 mg/dL) were more likely to die within 6 months, and the relative risk increased as the definition became more restrictive. WRF defined as ≥0.1 mg/dL was not statistically significantly associated with death (adjusted HR = 0.89, 95% confidence interval [CI] 0.50, 1.60). The trend of increasing relative risks for the more restrictive definitions, however, remained even after adjusting for age, LVEF, discharge creatinine, diabetes, and years of HF (HR = 1.19, 1.67, 1.91, and 2.90 for WRF defined as creatinine elevations ≥0.2, ≥0.3, ≥0.4, and ≥0.5 mg/dL, respectively) (Fig. 3).
When other in-hospital adverse events were forced into multivariable models, HRs were slightly lower than in the adjusted model without these forced covariates. However, the strong trend of increasing risks for more restrictive definitions was not different. For the less restrictive definitions of creatinine elevations ≥0.2 and ≥0.3 mg/dL, associations were not statistically significant (see Table 1). In these models, none of the adverse events were significant predictors of mortality after WRF was entered. When variables for angiotensin-converting enzyme inhibitors at discharge and for number of days to peak creatinine level were forced into the model as covariates, the results did not change. The interaction term for LVEF and WRF was not significant in these models. Varying the definitions of WRF affected the sensitivity and specificity for predicting mortality. Sensitivity decreased from 75% to 40% for increasingly restrictive definitions, and specificity increased from 25% to 75%. The receiver operator characteristic (ROC) curve showed no clear inflexion point. All definitions except a creatinine increase ≥0.1 mg/dL showed benefit over the null (diagonal line, representing the ROC curve for no added sensitivity and specificity benefit for different definitions) (see Table 1, Fig. 4).
When WRF was defined as a relative increase by at least 25% from baseline admission level to 2.0 mg/dL or higher, it was not significantly associated with mortality, and showed a high specificity (91%) but a low sensitivity (14%). Wald chi-square statistics indicated that only age and LVEF were stronger predictors of death after discharge, compared with any definition of WRF, except creatinine increase ≥0.1 mg/dL. Additionally, before adding WRF into the multivariable models, discharge creatinine and admission creatinine values were significant predictors of death. However, after WRF was added into these models, admission and discharge measures of creatinine were no longer significant. Moreover, Wald chi-square values suggested that WRF was more important than the single measures of creatinine (results not shown). Readmission No significant associations were found between all definitions of WRF and all-cause readmission, although after adjusting for age, sex, discharge creatinine, history of HF, diabetes, and previous HF admission, HRs also tended to increase with more restrictive definitions (HR = 0.71, 1.13, 1.33, 1.39, and 1.91 for WRF defined as creatinine elevations ≥0.1, ≥0.2, ≥0.3, ≥0.4, and ≥0.5 mg/dL, respectively). With more restrictive definitions, sensitivity decreased from 73% to 28%, whereas specificity increased from 23% to 80% (Table 2, see Fig. 4).
| | | | WRF Definitions | Readmitted (%) | Sensitivity (%) | Relative* Sensitivity Change (%) | Specificity (%) | Relative* Specificity Change (%) | Unadjusted HR (95% CI) | Adjusted** HR (95% CI) | Adjusted† HR (95% CI) | |
|---|
| WRF (≥0.1) | 46 | 73 | — | 23 | — | 0.71 | 0.76 | 0.82 | |
| No WRF | 51 | | | | | (0.32, 1.56) | (0.37, 1.55) | (0.61, 1.12) | |
| WRF (≥0.2) | 47 | 58 | −21 | 41 | +78 | 1.13 | 0.94 | 0.95 | |
| No WRF | 48 | | | | | (0.53, 2.41) | (0.48, 1.86) | (0.72, 1.24) | |
| WRF (≥0.3) | 49 | 47 | −18 | 57 | +24 | 1.33 | 1.08 | 1.06 | |
| No WRF | 45 | | | | | (0.63, 2.84) | (0.56, 2.11) | (0.80, 1.39) | |
| WRF (≥0.4) | 52 | 36 | −25 | 71 | +14 | 1.39 | 1.01 | 1.22 | |
| No WRF | 45 | | | | | (0.67, 2.93) | (0.52, 1.96) | (0.91, 1.63) | |
| WRF (≥0.5) | 56 | 28 | −20 | 80 | +14 | 1.91 | 1.18 | 1.42 | |
| No WRF | 44 | | | | | (0.88, 4.15) | (0.61, 2.26) | (1.03, 1.94) | |
| WRF (≥25%) | 67 | 13 | −53 | 94 | +17 | 1.46 | 1.14 | 1.34 | |
| No WRF | 45 | | | | | (0.50, 4.28) | (0.50, 2.61) | (0.88, 2.03) | |
| *Relative change calculated as percent loss or gain compared with previous definition of WRF. **Adjusted for age, sex, discharge creatinine, history of heart failure, diabetes, and previous heart failure admission. †Combined outcome of death or readmission, adjusted for same covariates. | | | | |
Using the definition of WRF as a relative creatinine increase by at least 25% to at least 2.0 mg/dL again demonstrated a high specificity (94%) and low sensitivity (14%). This highly restrictive definition was not a significant predictor of readmission in the adjusted model. Results for models did not change when a variable for number of days to peak creatinine level was added. Models for the combined outcome of death or readmissions and HF readmissions only showed similar results. For all proportional hazards models, proportionality assumptions were met, and residual plots also indicated adequate fit and no outliers. Functional decline or death WRF, according to any definition, was significantly associated with the combined endpoint of functional decline or death during the 6-month follow-up only for the most extreme elevation of 0.5 mg/dL (OR = 1.95, 95% CI 1.16, 3.28). WRF showed only nonsignificant trends of association with functional decline alone in survivors, suggesting that the strong association with mortality tended to drive significance and increased odds for the combined endpoint (OR = 0.84, 0.99, 1.02, 1.05, and 1.34 for WRF defined as creatinine elevations ≥0.1, ≥0.2, ≥0.3, ≥0.4, and ≥0.5 mg/dL, respectively). WRF defined as a relative increase by at least 25% from baseline admission level to a level of 2.0 mg/dL or higher was not significantly associated with functional decline (13.3%) (Table 3).
| | | | WRF Definition | Declinedor Died (%) | Sensitivity (%) | Relative* Sensitivity Change (%) | Specificity (%) | Relative* Specificity Change (%) | Unadjusted OR (95% CI) | Adjusted** OR (95% CI) | Adjusted† OR (95% CI) | |
|---|
| WRF (≥0.1) | 41 | 77 | — | 25 | — | 0.91 | 0.93 | 0.84 | |
| No WRF | 38 | | | | | (0.56, 1.46) | (0.57, 1.54) | (0.46, 1.55) | |
| WRF (≥0.2) | 41 | 60 | −22 | 42 | 68 | 1.07 | 1.05 | 0.99 | |
| No WRF | 39 | | | | | (0.71, 1.61) | (0.68, 1.62) | (0.59, 1.66) | |
| WRF (≥0.3) | 44 | 51 | −16 | 57 | +36 | 1.32 | 1.23 | 1.02 | |
| No WRF | 37 | | | | | (0.88, 1.98) | (0.79, 1.91) | (0.60, 1.73) | |
| WRF (≥0.4) | 46 | 38 | −25 | 70 | +24 | 1.42 | 1.39 | 1.05 | |
| No WRF | 37 | | | | | (0.93, 2.18) | (0.87, 2.34) | (0.60, 1.85) | |
| WRF (≥0.5) | 52 | 32 | −17 | 80 | +15 | 1.92 | 1.95 | 1.34 | |
| No WRF | 36 | | | | | (1.21, 3.06) | (1.16, 3.28) | (0.71, 2.52) | |
| WRF (≥25%) | 55 | 13 | −58 | 93 | +15 | 1.98 | 1.72 | 1.57 | |
| No WRF | 38 | | | | | (1.01, 3.89) | (0.83, 3.54) | (0.67, 3.72) | |
| *Relative change calculated as percent loss or gain compared with previous definition of WRF. **Adjusted for age, left ventricular ejection fraction, discharge creatinine, history of heart failure, diabetes, and baseline functional status. †Survivors only (332). Adjusted for age, left ventricular ejection fraction, discharge creatinine, history of heart failure, diabetes, and baseline functional status. | | | | |
Results for models did not change when a variable for number of days to peak creatinine level was added. The Hosmer and Lemeshow Goodness-of-Fit test indicated a satisfactory model fit, and the c statistics for these models exceeded 0.6. Other analyses To analyze whether discharge creatinine, baseline admission creatinine, and peak creatinine were important predictors of outcome or modified the significance of WRF definitions in predicting mortality, interaction terms and stratification models were tested. Discharge, admission, and peak creatinine alone were all significant predictors of mortality in unadjusted analysis. However, when a WRF term was added into models, these single measures of creatinine were no longer significant predictors, because WRF explained these associations in multivariable models. The interaction variables for admission, discharge, and peak creatinine and WRF definitions were not significant predictors of any outcome in multivariable models, and models by stratum confirmed these results, with HRs for WRF remaining consistent within strata. For example, using a definition of WRF as ≥0.3 mg/dL, the HRs for mortality across the subgroups with discharge creatinine levels of <1.0 mg/dL, 1.0 to <2.0, and ≥2.0 were 1.59, 1.49, and 1.46, respectively. Discussion WRF during an admission for HF, even if defined as a small elevation in creatinine, is an important, independent prognostic indicator of mortality after discharge. Additionally, the prognostic importance of WRF is consistent even in patients admitted with a normal baseline level creatinine, regardless of peak creatinine values, discharge values, and even after considering other major in-hospital adverse events, including cerebrovascular accident, myocardial infarction, and shock. Our data provide information across a spectrum of creatinine elevations and help to illuminate tradeoffs between different definitions of WRF. The choice of a cutoff, or “best” definition, of WRF has implications both for the number of patients identified as having WRF and for the magnitude of risk for mortality. When WRF was defined as creatinine elevations from ≥0.2 to ≥0.5 mg/dL, the percentage of patients with WRF who died increased for the most restrictive definitions, even though the actual number of patients defined as having WRF decreased. This increased specificity for predicting death, however, was achieved at the cost of decreased sensitivity. The effect size for the association with mortality increased with more restrictive definitions. The only definition that did not follow these trends was a relative increase of 25% or more from the admission value, which lacked sensitivity and showed a relatively smaller effect size for predicting mortality. WRF defined as an increase ≥0.1 mg/dL was not associated with poor outcomes. Applied in a clinical setting, our results suggest that using a definition of a creatinine elevation ≥0.5 mg/dL during admission would be highly specific in predicting mortality after hospital discharge, with an indication of about 2 to 3 times the risk of death for patients with WRF. In managing HF patients, however, clinicians may be inclined to use a more sensitive definition of WRF. In our data, even the group identified with a creatinine elevation ≥0.2 mg/dL experienced about a 20% increased risk of death. Overall, the strong relationships between any definition of WRF and adverse outcomes suggest that clinicians should monitor any changes in creatinine during hospital admission and potentially avoid any elevations if possible. Clinicians may, in fact, view WRF as a continuous variable with a continuous spectrum of risks. Identifying patients who have experienced any level of WRF during hospitalization, however transient or minor, may also improve monitoring for progression in symptomatic HF after discharge. Whether creatinine elevations during hospitalization can be prevented and whether their prevention would reduce the risk of death and readmission after discharge remains to be determined. Our results may be valuable for future investigations because cohort studies and clinical trials have sought an appropriate dichotomous cutpoint to define WRF. In other patient populations, definitions have focused on more extreme elevations in creatinine during hospitalization, from ≥0.5 mg/dL to elevations as high as ≥3 mg/dL during an admission.2, 3, 4, 5, 6, 7, 8, 9, 10 Larger studies on HF patients have used a cutpoint of ≥0.3 mg/dL,1, 21 and our findings support these lower levels of creatinine elevations as sufficiently inclusive, having a relatively high sensitivity and specificity combination compared with other definitions, although even lower thresholds could be useful in detecting poor outcomes. Our results on the prognostic value of WRF validate previous findings in a cohort of elderly HF patients, where WRF predicted significantly worse postdischarge mortality even after controlling for age, sex, and other important comorbidities.1 WRF has also been identified as an important predictor of in-hospital mortality and length of stay, and our findings now establish that WRF also affects longer-term outcomes.22 Some prior studies have focused on patients with baseline renal dysfunction and have overlooked patients with normal baseline admission creatinine who peak to only moderately abnormal levels. Analyses of the Studies of Left Ventricular Dysfunction (SOLVD) trials found that even moderate renal impairment at baseline predicted increased risks for long-term mortality, pump failure death, and death or hospitalization for HF.23 A recent analysis of SOLVD found that in patients with left ventricular dysfunction, even small decreases in glomerular filtration rate estimated from baseline serum creatinine predicted significant increases in risk for all-cause mortality.24 Similar results for mortality were reported for Second Prospective Randomized Study of Ibopamine on Mortality and Efficacy (PRIME-II) trial data.25 Yet in these studies and in other smaller cohort studies,26, 27 renal impairment was assessed by a single measure of creatinine clearance at baseline, not as a specific episode of WRF during hospital admission. For our models, WRF was a more powerful predictor of death than a single measure of renal dysfunction at discharge. Whether WRF is a marker or a true cause of severe or worsening HF is not clear. However, studies increasingly suggest that poor renal function could be directly responsible for adverse outcomes.10, 23, 25 In a study on inpatients who developed contrast media–associated renal failure, Levy et al. found that the significantly higher in-hospital mortality rates could not be explained by comorbidities alone.10 Dries et al. further suggested that renal insufficiency may have a causal role in the progression of left ventricular systolic dysfunction through retention of sodium and fluid, increases in cardiac filling pressures, and subsequent ventricular dilation.23 Finally, recent findings suggest that neurohormones affecting renal perfusion (angiotensin II, endothelin, nitric oxide, prostaglandins, natriuretic peptide, and vasopeptide inhibitors) may mediate poor outcomes.25, 28, 29, 30, 31, 32 An episode of WRF may indicate hyperreactivity to neurohormones (e.g., endothelin), which may induce poor outcomes beyond hospitalization. Thus although the measure of WRF in our study may represent acute dynamic changes in-hospital, this predictor may be better than a single baseline value because it could reflect patients' underlying neurohormonal profiles. If WRF directly contributes to an increased risk of adverse outcomes, then our findings underscore the importance of avoiding even small decrements in renal function during hospitalization. Creatinine measurement is not linearly related to the creatinine clearance. Thus it is surprising that risks of an absolute increase in creatinine remain consistent regardless of baseline creatinine. Nevertheless, trends in our data remain consistent by WRF definition and for different outcomes. At higher baseline creatinine levels, the same absolute decrement in creatinine reflects a smaller loss relative to the kidney function compared with lower baseline creatinine levels. Yet our results appear to indicate that an increase from any baseline level confers about the same risk. Moreover, using the definition of a 25% relative increase in creatinine to at least 2.0 mg/dL was not superior, as indicated by the lower risks compared with other relatively restrictive definitions and lack of statistical significance in predicting mortality. Also, this definition had lower sensitivity and specificity values for all outcomes. Limitations Because creatinine levels at only 3 time points were included in the abstraction (admission, peak, and discharge), characterization of trends were relatively simplistic in our study. Also, serum creatinine may be less accurate than other measures of renal performance, such as glomerular filtration rate.14 Second, though actual sensitivity and specificity values may appear to be low, the purpose of this analysis was not to establish whether WRF alone could optimally model risk of death and readmission in HF patients, but was rather to illuminate the tradeoffs between the various definitions of WRF presented. In our analysis, censoring for deaths may have slightly biased the proportional hazards model for readmissions because rates of deaths for patients with and without WRF were not the same. However, we also conducted a subsidiary analysis on the combined outcome of readmission and death, and results did not differ substantially. Finally, this was a single-center study with a modest sample size conducted at a University-affiliated hospital, so effect sizes of associations may not necessarily be generalizable to all HF patients. The sample size may not have been adequate to detect statistically significant associations between less restrictive definitions of WRF and outcomes, which is indicated by the relatively wider confidence intervals for readmissions and functional decline. However, given the consistent results for increasing risks with larger effect sizes for more restrictive definitions for all adverse outcomes, the consistency of the results underscores the implication that even small increases in creatinine are important, even if not statistically significant in our sample. Conclusion Although larger creatinine elevations predict the highest risk of poor outcomes, physicians monitoring HF patients should be aware that even minor changes in renal function could be significant, because an elevation as small as 0.2 mg/dL is still associated with adverse outcomes. This and more extreme elevations in creatinine consistently and powerfully predict death after discharge and are associated with other poor outcomes. A clinically appropriate definition of WRF undoubtedly also depends on the clinical context, which may determine the relative importance of identifying the greatest potential number of patients at risk versus the greatest potential effect size for early mortality. References 1.
1
Krumholz HM, Chen YT, Vaccarino V, Wang Y, Radford MJ, Bradford WD, et al.
Correlates and impact on outcomes of worsening renal function in patients ≥ 65 years of age with heart failure.
Am J Cardiol. 2000;85:1103–1110. 2.
2
Hou SH, Bushinsky DA, Wish JB, Cohen JJ, Harrington JT.
Hospital-acquired renal insufficiency: a prospective study.
Am J Med. 1983;74:243–248. Abstract |
Full-Text PDF (777 KB)
|
CrossRef
3.
3
Mangano CM, Diamondstone LS, Ramsay JG, Aggarwal A, Herskowitz A, Mangano DT.
Renal dysfunction after myocardial revascularization: risk factors, adverse outcomes, and hospital resource utilization. The Multicenter Study of Perioperative Ischemia Research Group.
Ann Intern Med. 1998;128:194–203. MEDLINE 4.
4
Shusterman N, Strom BL, Murray TG, Morrison G, West SL, Maislin G.
Risk factors and outcome of hospital-acquired acute renal failure. Clinical epidemiologic study.
Am J Med. 1987;83:65–71. Abstract |
Full-Text PDF (791 KB)
|
CrossRef
5.
5
Rasmussen HH, Ibels LS.
Acute renal failure. Multivariate analysis of causes and risk factors.
Am J Med. 1982;73:211–218. Abstract |
Full-Text PDF (931 KB)
|
CrossRef
6.
6
Wilkins RG, Faragher EB.
Acute renal failure in an intensive care unit: incidence, prediction and outcome.
Anaesthesia. 1983;38:628–634. MEDLINE |
CrossRef
7.
7
Bullock ML, Umen AJ, Finkelstein M, Keane WF.
The assessment of risk factors in 462 patients with acute renal failure.
Am J Kidney Dis. 1985;5:97–103. Abstract 8.
8
Groeneveld AB, Tran DD, van der Meulen J, Nauta JJ, Thijs LG.
Acute renal failure in the medical intensive care unit: predisposing, complicating factors and outcome.
Nephron. 1991;59:602–610. 9.
9
Gruberg L, Mintz GS, Mehran R, Gangas G, Lansky AJ, Kent KM, et al.
The prognostic implications of further renal function deterioration within 48 h of interventional coronary procedures in patients with pre-existent chronic renal insufficiency.
J Am Coll Cardiol. 2000;36:1542–1548. Abstract | Full Text |
Full-Text PDF (449 KB)
|
CrossRef
10.
10
Levy EM, Viscoli CM, Horwitz RI.
The effect of acute renal failure on mortality. A cohort analysis.
JAMA. 1996;275:1489–1494. MEDLINE 11.
11
Schocken DD, Arrieta MI, Leaverton PE, Ross EA.
Prevalence and mortality rate of congestive heart failure in the United States.
J Am Coll Cardiol. 1992;20:301–306. MEDLINE 12.
12
Harlan WR, Oberman A, Grimm R, Rosati RA.
Chronic congestive heart failure in coronary artery disease: clinical criteria.
Ann Intern Med. 1977;86:133–138. MEDLINE 13.
13
Vaccarino V, Kasl SV, Abramson J, Krumholz HM.
Depressive symptoms and risk of functional decline and death in patients with heart failure.
J Am Coll Cardiol. 2001;38:199–205. Abstract | Full Text |
Full-Text PDF (186 KB)
|
CrossRef
14.
14
Weinfeld MS, Chertow GM, Stevenson LW.
Aggravated renal dysfunction during intensive therapy for advanced chronic heart failure.
Am Heart J. 1999;138:285–290. Abstract | Full Text |
Full-Text PDF (70 KB)
|
CrossRef
15.
15
Katz S, Downs TD, Cash HR, Grotz RC.
Progress in development of the index of ADL.
Gerontologist. 1970;10:20–30. MEDLINE 16.
16
Smith LA, Branch LG, Scherr PA, Wetle T, Evans DA, Hebert L, et al.
Short-term variability of measures of physical function in older people.
J Am Geriatr Soc. 1990;38:993–998. MEDLINE 17.
17
Krumholz HM, Chen YT, Wang Y, Vaccarino V, Radford MJ, Horwitz RI.
Predictors of readmission among elderly survivors of admission with heart failure.
Am Heart J. 2000;139:72–77. Abstract | Full Text |
Full-Text PDF (55 KB)
|
CrossRef
18.
18
Pernenkil R, Vinson JM, Shah AS, Beckham V, Wittenberg C, Rich MW.
Course and prognosis in patients ≥70 years of age with congestive heart failure and normal versus abnormal left ventricular ejection fraction.
Am J Cardiol. 1997;79:216–249. Full Text |
Full-Text PDF (246 KB)
|
CrossRef
19.
19
Aronow WS, Ahn C, Kronzon I.
Prognosis of congestive heart failure after prior myocardial infarction in older men and women with abnormal versus normal left ventricular ejection fraction.
Am J Cardiol. 2000;85:1382–1384. Full Text |
Full-Text PDF (78 KB)
|
CrossRef
20.
20
Vasan RS, Larson MG, Benjamin EJ, Evans JC, Reiss CK, Levy D.
Congestive heart failure in subjects with normal versus reduced left ventricular ejection fraction: prevalence and mortality in a population-based cohort.
J Am Coll Cardiol. 1999;33:1948–1955. Abstract | Full Text |
Full-Text PDF (299 KB)
|
CrossRef
21.
21
Pitt B, Segal R, Martinez FA, Meurers G, Cowley AJ, Thomas I, et al.
Randomised trial of losartan versus captopril in patients over 65 with heart failure (Evaluation of Losartan in the Elderly Study, ELITE).
Lancet. 1997;349:747–752. Abstract | Full Text |
Full-Text PDF (89 KB)
|
CrossRef
22.
22
Gottlieb SS, Abraham WT, Butler J, Forman DE, Loh E, Massie BM, et al.
The prognostic importance of different definitions of worsening renal function in congestive heart failure.
J Card Fail. 2002;8:136–141. Abstract | Full Text |
Full-Text PDF (142 KB)
|
CrossRef
23.
23
Dries DL, Exner DV, Domanski MJ, Greenberg B, Stevenson LW.
The prognostic implications of renal insufficiency in asymptomatic and symptomatic patients with left ventricular systolic dysfunction.
J Am Coll Cardiol. 2000;35:681–689. Abstract | Full Text |
Full-Text PDF (197 KB)
|
CrossRef
24.
24
Al-Ahmad A, Rand WM, Manjunath G, Konstam MA, Salem DN, Levey AS, et al.
Reduced kidney function and anemia as risk factors for mortality in patients with left ventricular dysfunction.
J Am Coll Cardiol. 2001;38:955–962. Abstract | Full Text |
Full-Text PDF (287 KB)
|
CrossRef
25.
25
Hillege HL, Girbes AR, de Kam PJ, Boomsma F, de Zeeuw D, Charlesworth A, et al.
Renal function, neurohormonal activation, and survival in patients with chronic heart failure.
Circulation. 2000;102:203–210. 26.
26
Wang R, Mouliswar M, Denman S, Kleban M.
Mortality of the institutionalized old-old hospitalized with congestive heart failure.
Arch Intern Med. 1998;158:2464–2468. MEDLINE |
CrossRef
27.
27
Cowie MR, Wood DA, Coats AJ, Thompson SG, Suresh V, Poole-Wilson PA, et al.
Survival of patients with a new diagnosis of heart failure: a population based study.
Heart. 2000;83:505–510. 28.
28
Levin ER.
Endothelins.
N Engl J Med. 1995;333:356–363. MEDLINE |
CrossRef
29.
29
Schrier RW, Abraham WT.
Hormones and hemodynamics in heart failure.
N Engl J Med. 1999;341:577–585. MEDLINE |
CrossRef
30.
30
Goodfriend TL, Elliott ME, Catt KJ.
Angiotensin receptors and their antagonists.
N Engl J Med. 1996;334:1649–1654. MEDLINE |
CrossRef
31.
31
Friedrich EB, Muders F, Luchner A, Dietl O, Riegger GA, Elsner D.
Contribution of the endothelin system to the renal hypoperfusion associated with experimental congestive heart failure.
J Cardiovasc Pharmacol. 1999;34:612–617. MEDLINE |
CrossRef
32.
32
Abassi Z, Gurbanov K, Rubinstein I, Better OS, Hoffman A, Winaver J.
Regulation of intrarenal blood flow in experimental heart failure: role of endothelin and nitric oxide.
Am J Physiol. 1998;274:F766–F774. MEDLINE New Haven, Connecticut Atlanta, Georgia From the *Department of Medicine, Yale University School of Medicine, New Haven, Connecticut; †Division of Cardiology, Emory University School of Medicine, Atlanta, Georgia; ‡Section of Cardiovascular Medicine, Department of Medicine, Yale University School of Medicine, New Haven, Connecticut, §Section of Chronic Disease Epidemiology, Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut; ∥Center for Outcomes Research and Evaluation, Yale-New Haven Hospital, New Haven, Connecticut (current affiliation: McMaster University, Hamilton, Ontario, Canada); and ¶Section of Health Policy and Administration, Department of Epidemiology and Public Health, Yale University School of Medicine, New Haven, Connecticut ☆ Reprint requests to: Dr. Krumholz, Yale University School of Medicine, 333 Cedar Street, PO Box 208025, New Haven, CT 06520-8025. ☆☆ Dr. Watnick was a Robert Wood Johnson Clinical Scholar at Yale University during the time the work was conducted. She is currently affiliated with the Section of Nephrology, Oregon Health Sciences University and Portland VA Hospital. PII: S1071-9164(02)25403-2 doi:10.1054/jcaf.2003.3 © 2003 Published by Elsevier Inc. | |
|
FULL TEXT25403-2/journalimage?img=PIIS1071916402254032.gr1.lrg.gif&fig=&kwhquery=&issn=1071-9164&ishighres=false&allhighres=false&free=yes','journalimage');)
25403-2/journalimage?img=PIIS1071916402254032.gr2.lrg.gif&fig=&kwhquery=&issn=1071-9164&ishighres=false&allhighres=false&free=yes','journalimage');)
25403-2/journalimage?img=PIIS1071916402254032.gr3.lrg.gif&fig=&kwhquery=&issn=1071-9164&ishighres=false&allhighres=false&free=yes','journalimage');)
25403-2/journalimage?img=PIIS1071916402254032.gr4.lrg.gif&fig=fig4&kwhquery=&issn=1071-9164&ishighres=false&allhighres=false&free=yes','journalimage');)