BUN as an Independent Predictor of Post-Hospital-Discharge Mortality among Older Veterans

Abstract

Background/Objectives

An elevated blood urea nitrogen (BUN) in known to be an important prognostic indicator in patients with end-stage heart or kidney disease or certain other life-threatening illnesses. However, it is less certain as to whether an elevated BUN is an independent predictor of long-term mortality risk in less seriously ill patients. To address this issue, we examined the relationship between BUN and long-term mortality after adjusting for potential confounders and other indicators of health status/disease severity, in a select population of older medically stable Veterans.

Design

Long-term prospective cohort study.

Setting

Outpatient follow-up of patients discharged from a recuperative care and rehabilitation unit (RCRU) of a Department of Veterans Affairs Community Living Center.

Participants

383 older Veterans (mean age = 78.6±7.6 years, 98% male, and 87% white) discharged alive and in stable medical condition.

Measurements

At discharge, each subject completed a comprehensive assessment and was then monitored as an outpatient for up to 9.3 years. Associations between blood urea nitrogen at RCRU discharge and mortality were identified utilizing Cox proportional hazards (PH) regression analyses adjusting for conditions known to confound this relationship.

Results

Within the follow-up period, 255 subjects (67%) died. In the unadjusted Cox PH model, a BUN > 30 mg/dL was associated with a nearly 2-fold increased risk of mortality (hazard ratio 1.90, 95%CI 1.41 - 2.56). The association between BUN and long-term mortality remained highly significant after adjusting for potential confounders (hazard ratio 1.78, 95%CI 1.29 - 2.44).

Conclusion

Our findings support BUN levels as an independent predictor of long-term mortality in older, medically stable Veterans. An elevated BUN may be reflective of global health status rather than solely an indicator of the severity of acute illness or unstable chronic disease.

Introduction

Although the blood urea nitrogen (BUN) is often used as an indicator of renal function and/or hydration status, it is also known to be an independent predictor of mortality in numerous clinical settings and patient populations (1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15). Strong positive associations between BUN and the risk of subsequent mortality have been demonstrated among patients with serious acute illness such as myocardial infarction (3, 4), pneumonia (5), chronic obstructive pulmonary disease (COPD) (6), sepsis (7) and pancreatitis (8, 9), as well as chronically ill outpatients with advanced organ failure such as end-stage congestive heart failure (CHF) (10, 11), or chronic kidney disease (CKD) (13). BUN may also be elevated and have prognostic significance in patients taking high dose corticosteroids or other drugs such as diuretics (16, 17). However, it is less clear whether an elevated BUN has prognostic significance in individuals without such serious conditions. Using data from two large data sets, Hu et al demonstrated an inconsistent relationship between BUN and all-cause mortality in the general population (15). Their first set of analyses included data from presumably healthy older adults applying for life insurance. Among those who had a BUN within the standard reference range of 6–25 mg/dL, it was found that a BUN of 6 to 12 mg/dL was associated with a significantly higher long-term mortality than a BUN of 15 to 25 mg/dL. Risk did not increase with a BUN greater than the reference range. In a second set of analyses using NHANES III data, there was evidence of a ‘U-shaped' relationship between BUN and mortality. After adjusting for diagnoses, smoking status, age, and functional independence, a BUN below the reference range of 6 to 25 mg/ dL was associated with the highest mortality risk and a BUN in the ‘low normalrange of 6 to 10 mg/dL was associated with a significantly higher mortality than a BUN of 11 to 25 mg/dL. It was also found that a BUN above the reference range was associated with a significantly higher mortality than a BUN of 6 to 10 mg/dL. However, the paper did not state the number of observations with a BUN above the reference range or provide the range of values for BUN in this subgroup. Importantly, body weight data were not examined, yet the investigators surmised that subjects with a BUN below the reference range were at increased risk for mortality because they were probably malnourished. This and the other cited studies suggest that the BUN may be an indicator of the severity of any metabolic disarray that can lead to alterations in urea production, absorption, or excretion as well as other vital functions, all of which can have important prognostic implications. They also suggest that the relationship between BUN and mortality may be confounded by other disease processes. However, these inter-relationships are not well understood.

In prior studies examining nutrition and health, we noted a strong association between BUN and an increased risk of longterm mortality in older, clinically stable patients discharged from a recuperative care and rehabilitation unit (RCRU). (18, 19). Dichotomizing BUN at 30 mg/dL maximized its discriminatory power to identify those subjects at highest longterm mortality risk. However, like prior studies, we did not attempt to control for the various conditions described above that are known to confound the relationship between BUN and mortality when conducting these analyses. The purpose of the present study is to examine how adjusting for such confounders impacts the strength of the relationship between BUN and longterm mortality in the same population.

Methods

Participants

Over a four-year period, all admissions to the RCRU of a university-affiliated Veterans Affairs Medical Center were evaluated to determine study eligibility. A few patients were admitted to the unit each year for respite care. All respite patients were excluded from the study. Otherwise, all the patients admitted to the unit had undergone a comprehensive geriatric assessment by an interprofessional team of geriatric practitioners. Most (73%) were transfers from an acute care ward where they had been seen by the Geriatric Consult team. The remaining subjects had been admitted through the geriatric clinic after they had undergone a similar detailed evaluation in the outpatient Geriatric and Management unit. Patients who were assessed by the team to be medically stable, have a good prognosis for long-term survival, and to be in need of and likely to benefit from a stay on the recuperative care and rehabilitation unit (RCRU) were offered the option of admission to the unit. Patients with advanced disease (e.g., metastatic cancer, end-stage COPD, end-stage CHF) were excluded. Study subjects had to meet these admission criteria and be 65 years of age or older. Of all admissions during this time, 540 met all of these study eligibility criteria, and 446 agreed to enter the study. Written informed consent was obtained from all subjects or from a legal representative if the subject lacked adequate decision-making capacity. Of those enrolled, 4 participants died during their RCRU stay, 21 withdrew, two lacked laboratory data, and 36 were discharged in an unstable condition to an acute care hospital. The remaining 383 subjects who were discharged in a stable condition represent the population for the current study. The study was compliant with the regulations and ethical standards of the Declaration of Helsinki, Health Insurance Portability and Accountability Act, Department of Veterans Affairs, and the Institutional Review Board of the Central Arkansas Veterans Healthcare System.

The RCRU study and the study population are described in detail elsewhere (18, 19, 20). In brief, all participants received comprehensive nutritional, medical, psychosocial, and functional evaluations on a regularly scheduled basis during their stay and, if possible, on the day of discharge. In the event discharge occurred unexpectedly, the most recent data (which were usually obtained within 3 days prior to discharge) were designated as discharge data. After hospital discharge, these subjects were followed through phone calls, clinic visits, and medical record review until their death or 15 July 2015, whichever came first.

As stated above, BUN can be influenced by and have prognostic implications for a number of potentially serious chronic conditions such as CHF and COPD; it may also be elevated and have prognostic significance in patients who are undernourished or taking certain drugs such as corticosteroids or diuretics. Based on the medical evaluation, variables were created to indicate the presence or absence of each of these conditions to be investigated as potential confounders of the association between BUN and mortality. Body mass index (BMI) was included as the indicator of nutritional status. The Charlson Comorbidity Index (CCI) (21), functional status at RCRU discharge assessed using the Katz Index of ADLs (Katz Index) (22), and a diagnosis of hypertension (HTN), coronary heart disease (CAD), or diabetes mellitus (DM) were included in analyses as additional mortality risk indicators (21, 22).

Statistical Analysis

Descriptive statistics were used to summarize study participants characteristics. For the primary outcome, BUN was dichotomized as less than vs. greater than or equal to 30 mg/dL. Cox proportional hazards (Cox PH) regression was used to assess the association between BUN and long-term mortality using a series of different analytical models. For all analyses, survival time was the dependent variable, and all participants who were still alive were right censored at the end of follow-up. The first model included BUN as the only independent variable. A series of additional multivariable analyses were then performed. The additional analyses were conducted to determine if BUN remained significantly associated with mortality after controlling for different groups of covariates, which were grouped according to themes. The first group included demographic factors (age, sex, race) and smoking status; the second included all of the previously mentioned chronic conditions (CHF, COPD, CKD, and low body mass index) and discharge medications (corticosteroids, diuretics, angiotensin-converting enzyme inhibitors/ angiotensin receptor blockers, beta-blockers) that are known to potentially confound the relationship between BUN and mortality; and the third group included other mortality risk indicators (CCI, HTN, CAD, Katz Index, DM). Creatinine was not included in the analyses as it was highly correlated with the diagnosis of CKD, which is based on creatinine. We did not need to adjust for the presence of metastatic cancer as there were no patients with this diagnosis. To determine whether the association between BUN and mortality was primarily the result of the impact of acute illness, analyses were repeated in a sensitivity analysis excluding those with less than one year of follow-up (i.e., those who died or were lost to follow-up within one year of RCRU discharge).

Continuous variables were transformed into quartiles. After examining Kaplan-Meier survival curves and the results of univariable Cox PH regression analyses, quartiles were collapsed together to create a dichotomous variable when a natural binary cutpoint was indicated (i.e., the survival estimates for quartiles were similar to adjacent quartiles, and the overall p-value was smaller for a dichotomous versus quartile transformation). Since the Cox PH model assumes that the hazard function with respect to each independent variable remains proportional over time (i.e., the relative risk is constantly proportional), independent variables were individually entered into univariable Cox PH analysis that included an appropriate time-dependent interactive term. A non-significant time-dependent variable (i.e., parameter estimate was not significantly different from zero) was considered evidence for the validity of the proportional hazards assumption with respect to the given covariate (23, 24). All independent variables of interest met this assumption. Interaction terms of BUN with each of the covariates included in the final multivariable analysis were investigated. Analyses were conducted using SAS Enterprise Guide software (version 5.1, SAS Institute, Inc., Cary, NC). Significance was defined as a two-tailed p < 0.05.

Results

The 383 study subjects were primarily white (86.7%) men (98.4%) with a mean age of 78.6 + 7.6 years. Prior to discharge, they had resided on the RCRU for a median of 22 (interquartile range (IQR) 14-39) days. Although all the subjects were frail, there was a fairly wide range of functional disability within the population. This was quantified by the Katz Index of ADL scores, which ranged from zero (totally independent) to 12 (totally dependent) with a median of 1.0 (IQR 0 to 5.0). Upon discharge, 339 (88.5%) returned home and 44 (11.5%) were admitted to a nursing home. Additional characteristics of the study population are shown in Table 1 (21, 25).

Table 1.

Characteristics of Study Subjects

	Discharge BUNa
Variable	≥30 mg/dl (n=68)	<30 mg/dl (n=315)	P-value	All Subjects (n=383)
Demographics
Age, years, mean± SD	79.4 ± 7.6	78.5 ± 7.6	0.395	78.6 ± 7.6
Male, n (%)	67 (98.5)	310 (98.4)	1.000	377 (98)
White race, n (%)	59 (86.8)	273 (86.7)	0.983	332 (87)
Smoker, n (%)	6 (8.8)	47 (14.9)	0.187	53 (14)
Potential Confounders of Association between BUN and Mortality
A discharge diagnosis of:
Congestive Heart Failure, n (%)	31 (45.6)	74 (23.5)	<0.001	105 (27)
Chronic Kidney Disease, n (%)b	60 (88.2)	90 (28.6)	<0.001	150 (39)
Chronic Obstructive Pulmonary Disease, n (%)	25 (36.8)	116 (36.8)	1.000	141 (37)
Body Mass Index, admission, kg/m2, mean± SD	28.5 ± 6.5	25.8 ± 5.3	0.002	26.3 ± 5.6
At discharge taking a:
Corticosteroid, n (%)	4 (5.9)	28 (8.8)	0.420	32 (8)
Diuretic, n (%)	44 (64.7)	122 (38.7)	<0.001	166 (43)
ACE/ARB, n (%)c	42 (61.8)	165 (52.4)	0.160	207 (54)
Beta-Blocker, n (%)	46 (67.7)	183 (58.1)	0.145	229 (60)
Additional Disease Severity/ Mortality Risk Indicators at Discharge
CCI, median (IQR)d	4 (3-5)	2 (1-3)	<0.001	2 (1-4)
KATZ Index of ADLs, median (IQR)	1 (0-4)	1.5 (0-5)	0.733	1 (0-5)
A discharge diagnosis of:
Coronary Artery Disease, n (%)	46 (67.7)	136 (43.2)	<0.001	182 (48)
Hypertension, n (%)	62 (91.2)	272 (86.4)	0.280	334 (87)
Diabetes Mellitus, type 2, n (%)	44 (64.7)	95 (30.2)	<0.001	139 (36)

^a.BUN = blood urea nitrogen;

^b.Chronic Kidney Disease: established diagnosis based a three-month history of an estimated glomerular filtration rate (eGFR) below 60 mL/min per 1.73 m² (25);

^c.ACE/ARB = angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker;

^d.CCI = Charlson’s Co-Morbidity Index (21)

The median time from hospital discharge to study end was 7.0 (IQR 6.1 to 8.1) years, during which 255 (66.6%) of the participants were known to have died. Within the first year after discharge, 64 subjects (16.7%) died. In total, 27 subjects (7.1%) were lost to follow-up a median of 3.1 years (IQR 1.7 to 4.7 years) after discharge.

When discharge BUN was transformed into quartiles (as opposed to a dichotomous variable) before entry into the univariable Cox PH analysis, there was no mortality gradient across the lowest three quartiles; compared to the highest quartile for BUN, the hazard ratio for the first three quartiles ranged from 0.59 to 0.64 (p<0.01 for all). As the range of the highest BUN quartile was 28 to 79 mg/dL, ≥28 mg/dL was identified as a natural cutpoint for dichotomizing BUN in order to maximize its discriminatory power to identify those subjects at highest long-term mortality risk. However, there was essentially no difference between using a cutpoint of ≥28 mg/dL or a cutpoint of ≥30 mg/dL (range 30 to 79 mg/dL, n=68 vs. all others range 6 to 29 mg/dL, n=315) as used in our prior studies (18, 19, 20).

Similarly, the initial analyses of the continuous covariates according to quartiles of their distributions served to identify natural cutpoints for each that maximized its discriminatory power to identify those subjects at highest long-term mortality risk. This included age (highest quartile, range 85 to 99 years, n=92 vs. all others, range 64 to 84 years, n=291), BMI (lowest quartile, range 16.2 to 22.1 kg/m2, n=91 vs. all others, range 22.2 to 43.4 kg/m2, n=292), Katz Index (highest quartile, range 5 to 12, n=98, vs. all others, range 0 to 4, n=285), and CCI (greater than median, range 3 to 8, n=175 vs. all others, range 0 to 2, n=208).

In the unadjusted (univariable) Cox PH model of long-term mortality, a BUN > 30 mg/dL was associated with a nearly 2-fold increased risk of mortality (hazard ratio (HR) 1.90, 95%CI 1.41 - 2.56) (Figure 1). Adjusting for demographics, conditions known to influence both the BUN and mortality risk, and other mortality risk indicators did not substantively change this association (see Table 2). Substituting creatinine for the diagnosis of CKD in the Cox PH model statement did not appreciably change any of the results, regardless of whether creatinine was entered as a continuous or transformed (to either quartiles or dichotomized as described above) variable. Since none of the interaction terms (BUN x covariate) were found to be significant or to influence the results even when forced into the model, they were not included in the final model. The hazard ratios also remained significant when an elevated BUN was alternately defined per the hospital laboratory reference range as any value >25 gm/dL (fully adjusted model, HR 1.61, 95%CI 1.24 - 2.10). As shown in Table 2, the association between BUN and long-term mortality remained highly significant after all subjects that died (n=64) or were lost to follow-up (n=2) within 365 days after discharge were deleted from the database. The survival curve for BUN by quartiles in shown in Figure 2.

Figure 1.

Kaplan-Meier survival curves for the two groups of subjects stratified by discharge blood urea nitrogen (BUN) as indicated (see text for details)

Table 2.

Association between Blood Urea Nitrogen (BUN) and All-Cause Mortality during seven years of follow-up

	All-Cause Mortality (255 deaths in 383 participants)		Long-Term All-Cause Mortality (≥1 year)(191 deaths in 317 participants)b
Model	Hazard Ratio (95% CI)a	p-value	Hazard Ratio 95% CI)a	p-value
Unadjustedd	1.90 (1.41 – 2.56)	<.001	1.67 (1.15 – 2.41)	0.006
Adjusted for Demographicsc	2.06 (1.52 – 2.78)	<.001	1.85 (1.27 – 2.69)	0.001
Adjusted for Demographics & Confoundersd	1.87 (1.38 – 2.55)	<.001	1.65 (1.13 – 2.41)	0.010
Adjusted for Demographics, Confounders & Disease Severitye	1.78 (1.29 – 2.44)	<.001	1.53 (1.04 – 2.26)	0.033

^a.Hazard ratio for mortality and corresponding p-value associated with BUN ≥30 mg/dL based on Cox proportional hazards regression analyses;

^b.After excluding (i.e., deleting from the database) patients who were followed for less than 365 days subsequent to recuperative care and rehabilitation unit (RCRU) discharge (64 deaths, 2 lost to follow-up);

^c.Demographics = age, sex, race and smoking status;

^d.Confounders = chronic conditions (congestive heart failure, chronic obstructive pulmonary disease, chronic kidney disease, and low body mass index) and discharge medications (corticosteroids, diuretics, angiotensin-converting enzyme inhibitors/ angiotensin receptor blockers, beta-blockers) that are known to potentially confound the relationship between BUN and mortality (See text for details);

^e.Disease Severity: The Charlson Comorbidity Index, functional status at RCRU discharge assessed using the Katz Index of ADLs, and a diagnosis of hypertension, coronary heart disease, or diabetes mellitus were included in analyses as additional mortality risk indicators (See text for details).

Figure 2.

Kaplan-Meier Survival Curves for the study subjects stratified by quartiles of discharge blood urea nitrogen. The range of values for quartile 1 is 6 to 15 mg/dL, for quartile 2 16 to 20 mg/dL, for quartile 3 21 to 27 mg/dl, and for quartile 4 28 to 79 mg/dL

Discussion

In prior studies, we demonstrated a strong positive association between BUN and the subsequent risk of long-term mortality in older, medically stable patients discharged from an RCRU (18, 19). Although these were the first studies to clearly document such an association in this population, they left unanswered several important questions. Since the analyses did not fully adjust for all the conditions known to confound the relationship between BUN and mortality, it remained unclear whether the findings were simply a reflection of the fact that many of the patients in the sample had one or more of these conditions. To assess the importance of BUN as a general prognostic indicator among this population of older patients, further evaluation was needed. The purpose of this study was to examine this issue in more depth.

It is well established that BUN has important prognostic implications in patients with acute life-threatening inflammatory conditions (3, 5, 7, 8, 9), and relapsing or chronic progressive cardiac and kidney disease (10, 11, 13). It is also an important prognosticator in patients taking high dose corticosteroids or other drugs such as diuretics (16, 17). In all of these conditions, the association between BUN and mortality remains strong after controlling for creatinine. This may relate to the fact that these various conditions can independently affect the BUN through any of several mechanisms; this could include a decline in BUN excretion, an increase in urea absorption from the gut, and/or an increase in urea production. The decline in excretion can be independent of a change in GFR since it is often the result of poor renal perfusion with activation of the sympathetic nervous and renin-angiotensin-aldosterone systems. This activation leads to enhanced proximal tubular reabsorption of urea. Factors such as increased dietary protein or gastrointestinal bleeding can lead to increased absorption of urea from the gut. Increased urea production often indicates accelerated tissue catabolism. Regardless of the mechanism of the change, the rise in BUN in these conditions probably reflects the seriousness of the underlying condition or the dose of the given medication, thus explaining its prognostic significance.

Given that many of the conditions that can confound the association between BUN and mortality were highly prevalent within the study population, it was not surprising that BUN was found to be a strong predictor of mortality in the univariable analysis. Although adjusting for these potential confounders did not substantively change the univariable result, it did not eliminate the possibility that the apparent independent association between BUN and mortality remained highly significant because of confounding within a subset of patients with a particular diagnosis or other condition that affects BUN. To investigate this possibility, two-way interaction terms between BUN and covariates were examined in the final multivariable model, and none were significant. Although we recognize that the ability to control for all the variables confounding the association is beyond the reach of the study, these results suggest that BUN conveys important prognostic information beyond that contributed by the known confounders.

Excluding subjects followed less than one year after discharge from the analyses also did not substantively diminish the strength of the association between BUN and mortality. This suggests that it is not an acute unstable medical problem that accounts for the BUN-mortality association. It is also unlikely that the elevated BUN was the result of an acute or chronic gastrointestinal bleed, the addition of total parenteral or enteral nutrition, or acute renal or post-renal causes as there was no evidence of these conditions at the time of discharge in this closely monitored patient population. These results and the knowledge that all subjects in the present study were clinically stable at study entry suggests that BUN is a more global indicator of health status and is not just an indicator of acute severe illness or a destabilized chronic condition in this population. It also does not appear to be an indicator of illness burden or advanced disability as the association was also independent of other risk indicators such as the Katz Index and CCI, both of which were previously demonstrated to be significantly associated with long-term mortality risk in this population (18, 19). Instead, an elevated BUN at RCRU discharge may indicate the presence of other, more subtle metabolic disorders that cause alterations in urea production, absorption, or excretion. Given the high prevalence of chronic disease in this population, a combination of multiple factors is the most likely explanation. As shown in our prior investigations (19, 26, 27), many patients admitted to our RCRU had elevated inflammatory markers suggestive of ongoing dysregulated subclinical inflammation. This could cause accelerated tissue catabolism that could lead to greater BUN production, a net negative nitrogen balance, and a decline in lean body mass and physiologic function resulting in a higher mortality risk.

In a prior study, Hu et al found evidence of a possible ‘U-shaped' relationship between BUN and mortality (15). They also found a mortality gradient across BUN within its reference range. We did not confirm either of these findings. Instead, we found that there was an apparent threshold effect for BUN with mortality only increasing when BUN was greater than 25 to 30 mg/dL. However, we were not able to assess the relationship between a low BUN and mortality as there were no subjects in this study with a BUN below the reference range as there was in the Hu study. Our failure to identify a mortality gradient across BUN within its reference range may relate to differences in study populations or the smaller sample size of our study.

Strengths of Study

The study has a number of strengths including its prospective design, an ample sample size, long-term subject follow-up, and comprehensive assessment with very little missing data. Due to the richness of the subject data, tight control for potentially confounding variables was possible, a feature that is unique to this paper. The study population represented medically stable, though clinically complex older patients (as indicated by the high prevalence of comorbid conditions) for which risk stratification may prove useful clinically. The primary outcome was all-cause mortality, which is considered an unbiased and clinically relevant outcome in long-term observational studies (28, 29).

Study Limitations

The study also had several noteworthy limitations. Given its observational design, it is possible that both identified and unidentified confounders may have influenced the outcomes despite attempts to control for several known risk factors through multivariable analyses. Further the sample is limited by overrepresentation of older Caucasian males, who were selected with non-probability sampling in a single center study. The study used only baseline assessment parameters obtained at single point in time (discharge); repeated clinical evaluations during follow-up may provide additional information to better model changes in risk over time and clinical opportunities to ameliorate risk. Another potential concern is the use of transformed (dichotomous) component measures. Although these transformed variables may have made the risk scores more applicable to this population and simplified reporting, they may have also caused a loss in some information and limited generalizability. Further validation of the risk scores may be useful. Although the investigation controlled for certain diagnoses and medications, we did not look at medication dosages or other diagnosis-specific indicators of disease severity. These classifications may provide additional elucidation of the actual meaning of BUN values in this population. However, it would be a challenge to assign certain diagnosis-specific indicators of disease severity (e.g., New York Heart Association classification for heart failure) given that many subjects were deconditioned and in a recuperative phase of an illness; ascribing symptoms (such as breathlessness) to a given diagnosis was problematic at best. Mortality was determined by accessing patient information in the electronic health record or by direct follow-up, with 7.1% (n=27) lost in follow-up. However, we believe it unlikely that using the Social Security Index would significantly change these results given the follow-up in this older population.

Conclusions

In this study, which focuses on a unique population of high-risk older Veterans being discharged from an RCRU, an elevated BUN provided prognostic information after adjustments for baseline clinical characteristics and potential confounders. Although the reasons for this association are probably complex and varied, BUN appears to be an indicator of the overall metabolic health of these older individuals. Additional prospective analyses of this relationship may help clarify the related physiology.

Author Contributions

DHS, MMB, and PKR conceived and designed the study; MMB and DHS oversaw the conduct of the study; DHS, MMB, SCS, PKR, and SYL analyzed the data; and, SCS and DHS wrote the paper.

Conflicts of Interest

The authors declare no conflict of interest.

Sponsors Role

The funding sponsors had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

Ethical standard

The study was compliant with the regulations and ethical standards of the Declaration of Helsinki, Health Insurance Portability and Accountability Act, Department of Veterans Affairs, and the Institutional Review Board of the Central Arkansas Veterans Healthcare System.