Article ID: CJ-24-0502
Background: The association between blood urea nitrogen (BUN) levels and incident heart failure (HF) in the general population is still unclear.
Methods and Results: We assessed the association of BUN level with incident HF in 14,167 ARIC participants without a history of HF at baseline (1987–1989) (mean age 54.1 years, 54.4% female, 25.2% Black). BUN levels (mg/dL) were divided into quartiles, with the highest quartile further divided into tertiles (Q1 ≤13, Q2 13–15, Q3 15–17, Q4a 17–19, Q4b 19–21, Q4c >21). HF events were identified through to December 31, 2019, using diagnostic codes on discharge records or death certificates. Hazard ratios (HRs) were estimated using multivariable Cox models. During a median follow-up of 26.2 years, 3,482 participants developed HF (incidence rate 10.7 per 1,000 person-years). In a multivariable Cox model adjusted for sociodemographic variables, the highest BUN quartile (Q4) had a HR of 1.19 (95% confidence interval [CI] 1.09, 1.31) compared with Q1. HRs for Q4a, Q4b, and Q4c were 1.14 (95% CI 1.02, 1.28), 1.11 (0.96, 1.28), and 1.42 (1.22, 1.63), respectively. After further adjustment for clinical factors, the association remained significant for Q4c (HR 1.23 [1.06, 1.43]). Associations were consistent across demographic and clinical subgroups.
Conclusions: In this community-based cohort, higher BUN levels were significantly associated with incident HF. BUN, routinely measured in clinical care, may help identify individuals at risk of HF.
Blood urea nitrogen (BUN) is routinely assessed in clinical practice and often included in a metabolic panel, together with other metabolic markers such as creatinine and electrolytes.1 Urea nitrogen is a waste product from protein breakdown and is primarily eliminated from the body through the kidneys. Thus, impaired kidney function is a major cause of elevated BUN levels. In addition, elevated BUN levels are also seen in states that increase protein catabolism, such as dehydration, high protein intake, gastrointestinal bleeding, or use of certain medications.2
Heart failure (HF) is a serious medical condition that can disrupt normal blood flow and lead to decreased kidney function, resulting in renal hypoperfusion and fluid retention in the body3 and causing a variety of metabolic imbalances, including elevated BUN levels.4–11 Importantly, among patients with acute decompensated HF, higher levels of BUN have been associated with adverse outcomes such as death and HF readmission. For example, a large USA study with >50,000 patients with HF identified BUN as one of the most potent prognostic predictors, together with systolic blood pressure (SBP) and creatinine level.12
However, little is known about the association of BUN level with the development of HF in the general population. To the best of our knowledge, there is only one study that explored this research question and it found no association.13 However, the study had a few important caveats, such as not adjusting for kidney function, only investigating a single racial group (i.e., Chinese people), and having a short follow-up of <4 years.
Therefore, we examined the association of BUN with the subsequent risk of HF using data from a biracial community-based cohort, the Atherosclerosis Risk in Communities (ARIC) Study, with >3 decades of follow-up. If BUN is robustly associated with incident HF risk, our results would have clinical implications, because BUN is widely measured in clinical settings but not specifically used to classify HF risk. Notably, the value of HF risk prediction is greater than ever, given the availability of new medications reducing the risk of HF.14–16
The ARIC Study is a prospective cohort study, enrolling a total of 15,792 participants aged 45–64 years between 1987 and 1989 (visit 1) from four USA communities: Forsyth County, NC; Jackson, MS; the northwest suburbs of Minneapolis, MN; and Washington County, MD. In the present study, we excluded participants with missing BUN values (n=150) or covariates (n=422), those with HF (n=1,009), and those with self-identified race other than Black or White (n=44), leaving 14,167 participants in our analytic sample (Figure 1). All participants provided written informed consent (including the use of medical records). The Institutional Review Board at the ARIC Study centers approved the study protocol, which was conducted in accordance with the Declaration of Helsinki and the ethical standards of the responsible committee on human experimentation.
Flowchart of participants’ inclusion and exclusion. ARIC, Atherosclerosis Risk in Communities; BUN, blood urea nitrogen; HF, heart failure.
Exposure
The exposure was BUN levels measured in the plasma sample at visit 1, using the DART BUN reagent with an enzymatic conductivity rate method based on the modification of the Talke and Schubert method.17 Urea in the sample was broken down into ammonia by urease and then reacted with glutamate dehydrogenase (GLD). Reaction with GLD led to a decrease in the amount of nicotinamide adenine dinucleotide (NADH). Because the abundance of NADH can be quantified by measuring the degree of absorbance at 340 nm, the concentration of urea nitrogen is calculated as a function of decreasing rate of NADH abundance.
CovariatesCovariates were determined a priori based on previous literature such as the Framingham risk score for composite cardiovascular outcomes including HF:12,18 age, sex, race, ARIC center, education level, total cholesterol level, high-density lipoprotein (HDL) cholesterol, SBP, smoking status, diabetes mellitus, body mass index (BMI), antihypertensive medication, lipid-lowering medication, history of coronary artery disease (CAD), drinking status, and estimated glomerular filtration rate (eGFR). Age, race, sex, education level (less than high school, high school but not college, and college or higher), drinking status (current, former and never), and smoking status (current, former, and never) were based on self-reports. Participants were asked to bring containers of all prescribed medications in the past 2 weeks, and trained staff recorded relevant medications using that information or Medi-Span Generic Product Identifier code. BMI was calculated as weight in kilograms divided by the square of height in meters. Blood pressure was measured three times using a sphygmomanometer after 5-min rest and the average of the 2nd and 3rd measurements were used in the analysis.19 Total cholesterol and HDL cholesterol levels in serum were measured using an enzymatic assay.20 eGFR was calculated using the CKD-EPI creatinine equation.21 Creatinine level was measured by a modified Jaffe method.22 Diabetes mellitus was defined as fasting blood glucose ≥126 mg/dL, non-fasting blood glucose ≥200 mg/dL, self-reported history of physician diagnosis, or use of antidiabetic medications. History of CAD was defined by a self-reported history of myocardial infarction or coronary revascularization or silent myocardial infarction based on ECG.
OutcomesOur outcome of interest was incident HF. In our main analysis, we identified 3,481 HF cases through discharge diagnosis of HF using the International Classification of Diseases ninth revision (ICD-9) 428 or 10th revision (ICD-10) I50 at any diagnostic position. The ARIC Study investigators started to adjudicate acute decompensated HF events from 2005 onward. As a secondary analysis, we used HF events that were ascertained through ICD codes up to the end of 2004 (1,453 cases) and adjudicated definite or probable HF cases starting from 2005 onward (1,071 cases).
Statistical AnalysisBaseline characteristics are summarized across quartiles of BUN and compared using t-tests for continuous variables and chi-square tests for categorical variables. Incidence rates are calculated using Poisson regression. The Kaplan-Meier method was used to estimate the cumulative incidence of HF across quartiles of BUN. Log-rank test was used to statistically assess between-group differences in the incidence of HF.
We used Cox proportional hazards models to estimate the hazard ratios (HRs) of incident HF. We ran 3 sequential models to incorporate potential confounders. Model 1 adjusted for age, sex, race, and ARIC center. Model 2 additionally accounted for education level, drinking status, smoking status, BMI, SBP, antihypertensive medication, lipid-lowering medication, diabetes, total cholesterol, HDL cholesterol, and history of CAD. Model 3 was further adjusted for eGFR. Unless mentioned otherwise, the lowest quartile (Q1) served as the reference.
We performed several sensitivity analyses. First, as noted above, we repeated the analysis using HF cases combining ICD-based cases and adjudicated cases. Second, we dichotomized BUN by a clinical threshold of 24 mg/dL.23 Third, given the possibility that low hemoglobin levels due to bleeding or use of diuretics may affect both BUN levels and risk of HF, we tested a model in which hemoglobin level or the use of diuretics was added to Model 3. Finally, we performed subgroup analysis by age (≥ vs. <54 years [median]), race (White vs. Black), sex (male vs. female), BMI (≥ vs. < 26.7 kg/m2 [median]), drinking status (current vs. former/never), smoking status (current vs. former/never), diabetes (yes vs. no), antihypertensive medication use (yes vs. no), and history of CAD (yes vs. no). Statistical significance of interaction was based on the likelihood ratio test. All analyses were performed using the RStudio version 4.2.3 (Posit Software, PBC). A two-sided P value <0.05 was considered statistically significant.
Of the 14,167 participants in this study, the mean age (SD) was 54.1 (5.8) years, 54.4% were female, and 25.2% self-identified as Black (Table 1). Participants with higher levels of BUN tended to be older, White, and male. They were also more likely to have lower eGFR, diabetes, a history of CAD, and to be on antihypertensive and lipid-lowering medications.
Baseline Characteristicsa According to Quartiles of BUN Level
| BUN level (mg/dL) | P valueb | ||||
|---|---|---|---|---|---|
| Q1 (≤13 mg/dL) | Q2 (13–15 mg/dL) | Q3 (15–17 mg/dL) | Q4 (>17 mg/dL) | ||
| Number at risk | 4,841 | 3,383 | 2,701 | 3,242 | |
| BUN | 11.38 (1.6) | 14.51 (0.5) | 16.47 (0.5) | 20.54 (4.4) | <0.001 |
| Age (years) | 52.9 (5.6) | 53.9 (5.7) | 54.7 (5.7) | 55.6 (5.7) | <0.001 |
| Male | 1,585 (32.7) | 1,492 (44.1) | 1,398 (51.8) | 1,983 (61.2) | <0.001 |
| Black | 1,619 (33.4) | 848 (25.1) | 574 (21.3) | 528 (16.3) | <0.001 |
| Location of medical center | <0.001 | ||||
| Forsyth County, NC | 1,351 (27.9) | 916 (27.1) | 670 (24.8) | 766 (23.6) | |
| Jackson, MS | 1,431 (29.6) | 734 (21.7) | 490 (18.1) | 455 (14.0) | |
| Minneapolis, MN | 1,150 (23.8) | 864 (25.5) | 795 (29.5) | 968 (29.9) | |
| Washington County, MD | 909 (18.8) | 871 (25.7) | 746 (27.6) | 1,053 (32.5) | |
| Education level | 0.291 | ||||
| Less than high school | 1,072 (22.2) | 751 (22.2) | 602 (22.3) | 773 (23.8) | |
| High school but not college | 2,031 (42.0) | 1,407 (41.6) | 1,120 (41.5) | 1,273 (39.3) | |
| College or higher | 1,738 (35.9) | 1,225 (36.2) | 979 (36.2) | 1,196 (36.9) | |
| BMI | 27.4 (5.6) | 27.6 (5.3) | 27.5 (5.0) | 27.6 (4.8) | 0.063 |
| SBP (mmHg) | 120.8 (19.2) | 120.8 (18.6) | 120.9 (18.2) | 121.1 (18.3) | 0.901 |
| DBP (mmHg) | 74.9 (11.7) | 73.4 (10.9) | 73.2 (10.8) | 73.3 (10.8) | 0.012 |
| Comorbidities | |||||
| Diabetes | 481 (9.9) | 332 (9.8) | 283 (10.5) | 444 (13.7) | <0.001 |
| Previous CAD history | 133 (2.7) | 139 (4.1) | 114 (4.2) | 202 (6.2) | <0.001 |
| Medications | |||||
| Anti-HTN medication | 1,087 (22.6) | 876 (25.9) | 736 (27.3) | 1,134 (35.0) | <0.001 |
| Lipid-lowering medication | 94 (1.9) | 93 (2.7) | 87 (3.2) | 118 (3.6) | <0.001 |
| Total cholesterol (mmol/L) | 5.5 (1.1) | 5.5 (1.1) | 5.6 (1.1) | 5.6 (1.1) | 0.002 |
| HDL cholesterol (mmol/L) | 1.39 (0.45) | 1.35 (0.45) | 1.32 (0.44) | 1.27 (0.41) | <0.001 |
| eGFR (mL/min/1.73 m2) | 108.9 (13.6) | 103.7 (13.0) | 100.1 (13.5) | 93.6 (17.2) | <0.001 |
| Hemoglobin g/dL | 13.7 (1.4) | 13.9 (1.4) | 14.0 (1.3) | 14.1 (1.4) | <0.001 |
| Smoking status | <0.001 | ||||
| Current smoker | 1,579 (32.6) | 889 (26.3) | 571 (21.1) | 617 (19.0) | |
| Former smoker | 1,302 (26.9) | 1,053 (31.1) | 980 (36.3) | 1,263 (39.0) | |
| Never smoker | 1,960 (40.5) | 1,441 (42.6) | 1,150 (42.6) | 1,362 (42.0) | |
| Drinking status | <0.001 | ||||
| Current drinker | 2,627 (54.3) | 1,907 (56.4) | 1,621 (60.0) | 1,923 (59.3) | |
| Former drinker | 878 (18.1) | 625 (18.5) | 479 (17.7) | 640 (19.7) | |
| Never drinker | 1,336 (27.6) | 851 (25.2) | 601 (22.3) | 679 (20.9) | |
aFor continuous variables, the column shows mean (SD); for categorized variables, the column shows number (%). bAll P values from t-test for continuous variables and chi-square for categorical variables across quartiles of BUN. BMI, body mass index; BUN, blood urea nitrogen; CAD, coronary artery disease; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein; HTN, hypertension; MD, Maryland; MN, Minnesota; MS, Mississippi; NC, North Carolina; SBP, systolic blood pressure; SD, standard deviation.
Association Between BUN and Incident HF
During a median follow-up of 26.2 years, there were 3,482 (24.6%) incident HF cases (crude incidence rate, 10.7 per 1,000 person-years). Crude incidence rate and adjusted incidence rate were higher in Q4 vs. Q1 (e.g., 16.9 vs. 10.8 per 1,000 person-years in Model 3) (Supplementary Table 1). The cumulative incidence of HF was higher along with BUN levels in a graded manner (Figure 2, log-rank test P<0.0001). For example, the 30-year cumulative incidence was 29.2% in Q4 and 21.4% in Q1.
Cumulative incidence of heart failure across quartiles of blood urea nitrogen levels.
In a multivariable Cox model adjusting for demographics, the highest quartile of BUN vs. the lowest quartile was associated with a significantly higher risk of HF (HR=1.19 [95% confidence interval (CI) 1.09, 1.31]; Model 1 in Table 2). When we subdivided Q4 into tertiles, Q4c demonstrated a HR of 1.42 (1.22, 1.63) (Model 1 in Table 2). When we further adjusted for major clinical factors, the associations were very similar (e.g., HR 1.16 [95% CI 1.06, 1.28] in Q4 and 1.35 [95% CI 1.17, 1.56] in Q4c, Model 2 in Table 2). Further adjustment for eGFR attenuated the results, but the association for Q4c remained statistically significant (HR 1.23 [95% CI 1.06, 1.43], Model 3 in Table 2). When we modeled BUN continuously (per 1 SD), we observed similar significant associations across Models 1–3. When we stratified by follow-up (< vs. ≥15 years), the association between BUN and HF was more evident in the earlier follow-up period (Supplementary Table 2).
Adjusted HRs of Incident HF Across BUN Categories
| BUN | HR (95% CI) | HR (95% CI) | HR (95% CI) | |||
|---|---|---|---|---|---|---|
| Model 1a | P valued | Model 2b | P valued | Model3c | P valued | |
| Q1 (≤13 mg/dL) | Ref. | – | Ref. | – | Ref. | – |
| Q2 (13–15 mg/dL) | 1.06 (0.96, 1.16) | 0.227 | 1.05 (0.96, 1.15) | 0.314 | 1.03 (0.94, 1.13) | 0.514 |
| Q3 (15–17 mg/dL) | 1.00 (0.91, 1.11) | 0.971 | 1.04 (0.94, 1.15) | 0.436 | 1.00 (0.91, 1.11) | 0.915 |
| Q4 (>17 mg/dL) | 1.19 (1.09, 1.31) | <0.001 | 1.16 (1.06, 1.28) | 0.002 | 1.10 (0.99, 1.21) | 0.060 |
| Q4a (17–19 mg/dL) | 1.14 (1.02, 1.28) | 0.020 | 1.11 (0.99, 1.24) | 0.073 | 1.07 (0.95, 1.20) | 0.275 |
| Q4b (19–21 mg/dL) | 1.11 (0.96, 1.28) | 0.163 | 1.12 (0.96, 1.29) | 0.144 | 1.06 (0.92, 1.23) | 0.422 |
| Q4c (>21 mg/dL) | 1.42 (1.22, 1.63) | <0.001 | 1.35 (1.17, 1.56) | <0.001 | 1.23 (1.06, 1.43) | 0.007 |
| 1-SD increment | 1.13 (1.09, 1.16) | <0.001 | 1.11 (1.07, 1.14) | <0.001 | 1.10 (1.06, 1.14) | <0.001 |
aModel 1 adjusted for age, sex, race, and ARIC center. bModel 2 further adjusted for education, smoking, drinking, BMI, SBP, antihypertension medication, lipid-lowering medication, diabetes, total cholesterol, HDL cholesterol, and previous CAD history in addition to Model 1. cModel 3 further adjusted for eGFR in addition to Model 2. dP value is calculated by Wald test. ARIC, Atherosclerosis Risk in Communities; HF, heart failure; HR, hazard ratio. Other abbreviationsa as in Table 1.
Sensitivity Analysis
When we used HF events that were ascertained through ICD codes (before 2005) and physician’s adjudication (from 2005 onward), the results were similar to those from the primary analysis. For example, the HR of HF was 1.13 (1.01, 1.27) in Q4 and 1.19 (1.00, 1.43) in Q4c (Supplementary Table 3) vs. Q1, with significant associations for BUN as a continuous variable in all three models. When we dichotomized BUN, elevated BUN > 24 vs. ≤ 24 mg/dL showed significant HRs in Models 1–3 (e.g., 1.78, [95% CI 1.46, 2.16], Model 3 in Supplementary Table 4). The addition of hemoglobin levels or the use of diuretics to Model 3 did not change the association (Models 4 and 5, respectively, in Supplementary Table 5).
Finally, the associations were largely consistent across the subgroups tested (Figure 3). For example, the HRs for BUN >24 vs. ≤24 mg/dL were respectively 1.82 [95% CI 1.46, 2.25] and 1.72 [95% CI 1.07, 2.74] in age groups ≥54 and <54 years, and 1.93 [95% CI 1.36, 2.72] and 1.70 [95% CI 1.34, 2.16] in Black and White participants. A statistically significant interaction was seen only for the use of antihypertensive medication (P for interaction=0.04), where the HR was greater in participants taking this class of medication than those not taking, although a positive association was observed in both groups.
Hazard ratio of incident heart failure across quartiles of blood urea nitrogen levels among subgroups. BMI, body mass index; CAD, coronary artery disease; CI, confidence interval.
In this community-based cohort study with follow-up over 30 years, higher BUN levels were significantly associated with incident HF after adjusting for established cardiovascular risk factors. The association was somewhat attenuated when we accounted for eGFR, but remained largely consistent. A series of sensitivity analyses confirmed the robust associations between BUN and incident HF. For example, the HR of incident HF was almost 2 when we contrasted elevated BUN (>24 mg/dL) vs. normal BUN (≤24 mg/dL) even in our fully adjusted model.
Although we observed that elevated BUN levels were associated with incident HF in the general population, as noted in the Introduction, a previous Chinese study13 did not find a significant association between BUN and incident HF. The discrepancy between that study and ours may be due to different study populations and designs (e.g., ~4 years of follow-up in the Chinese study vs. ~30 years in our study). Of note, Jiang et al. used the same Chinese cohort and reported that higher BUN levels were associated with increased risk of CAD.4
Although we are not fully sure about the exact mechanisms linking BUN to the incidence of HF, there are a few plausible mechanisms. For example, elevated BUN levels could represent reduced kidney function, which is a known potent risk factor for HF.24 However, we should acknowledge that the association remained significant even after adjusting for eGFR in our study. BUN levels are also elevated in hypercatabolic states, a condition that may activate neurohormonal, pro-inflammatory, and mammalian target of rapamycin (mTOR) pathways25 and may contribute to the development of HF.26,27 Also, elevated BUN levels may also activate the sympathetic nervous system and the renin-angiotensin-aldosterone system,28 which are involved in the pathophysiology of HF by affecting renal blood flow and contributing to fluid retention. Nonetheless, our findings need to be replicated in other settings, and if so, future studies are needed to explore the pathophysiological mechanisms behind the BUN-HF relationship.
Our study has several clinical implications. First, BUN is a common clinical test and is often measured together with creatinine to evaluate kidney function. However, kidney function is almost exclusively assessed by eGFR calculated with creatinine, and BUN may attract less attention in clinical settings other than for specific occasions (e.g., bleeding and acute kidney injury). Thus, when BUN data are available, our findings suggest it may be a potential biomarker of individuals at risk of HF. Such utilization may be particularly relevant in resource-limited settings where measurement of other established biomarkers such as natriuretic peptides is not readily or widely available. Again, this concept should be tested in future studies.
Although our study showed a robust association of BUN with future risk of HF, whether lowering BUN levels could reduce the risk of HF is a topic for future investigation. In this context, we need to acknowledge that tolvaptan, a medication reducing BUN levels,29 did not improve prognosis of patients with HF.30 Nonetheless, as discussed earlier, BUN may be useful to identify individuals at risk of HF and lead to early intervention of established risk factors of HF (e.g., hypertension). This concept also should be evaluated in future studies.
Study LimitationsFirst, the main analysis used diagnostic codes to ascertain HF events, which might have led to some misclassification of cases.31 However, combining code-based HF events prior to 2005 and adjudicated HF events after 2005 showed similar results. Second, we only captured hospitalized HF, and thus some mild HF cases may have been missed. Third, we used baseline BUN levels at a single time point. A prior study reported that persistently high BUN levels, not a transient elevation, predicted poor prognosis in patients with HF.3 Fourth, we only included Black and White participants in our study, so our findings may not be generalizable to other racial/ethnic groups. Finally, as in any observational studies, we cannot deny the possibility of residual confounding.
Study StrengthsFirst, our study is one of the largest, allowing us to examine several important demographic and clinical subgroups. Second, we uniquely had a long follow-up over 3 decades. Third, we adjusted for many relevant potential confounders including eGFR. Fourth, we conducted a series of sensitivity analyses and confirmed the robustness and consistency of our findings.
In conclusion, this study found a significant association between elevated BUN levels and incident HF. The association remained significant even after adjusting for confounders, including eGFR. These findings suggest that BUN could be a potential biomarker of individual at risk of developing HF. However, future studies are needed to replicate our findings, and to test the clinical utility of BUN to predict future risk of HF and examine potential causal pathways linking BUN to HF.
We thank the staff and participants of The Atherosclerosis Risk in Communities Study for their important contributions.
The Atherosclerosis Risk in Communities (ARIC) study has been funded in whole or in part with federal funds from the National Heart, Lung, and Blood Institute, National Institutes of Health, Department of Health and Human Services, under contract nos. (HHSN268201700001I, HHSN268201700002I, HHSN 268201700003I, HHSN268201700004I, HHSN268201700005I).
The authors declare no conflicts of interest.
The ARIC study was approved by the Johns Hopkins Medicine IRB-3, with reference number IRB00311861, IRB00311999.
The deidentified participant data will be shared on a request basis. Please contact the corresponding author to request data sharing.
Please find supplementary file(s);
https://doi.org/10.1253/circj.CJ-24-0502
