Saliva Urea Nitrogen Continuously Reflects Blood Urea Nitrogen after Acute Kidney Injury Diagnosis and Management: Longitudinal Observational Data from a Collaborative, International, Prospective, Multicenter Study

The global ‘0by25' initiative recently launched by the International Society of Nephrology (ISN) aims to eradicate preventable deaths resulting from acute kidney injury (AKI) [1]. Along with an increased community attention, AKI is being increasingly recognized as a challenge to healthcare, particularly in regions with limited medical infrastructure [1,2]. While AKI in developing countries is generally found in an older and sickly population, often hospitalized or in an intensive care unit, in developing countries AKI is more often community-acquired, and caused by potentially avoidable and modifiable conditions such as diarrheal diseases, malaria or obstetric complications [1,2]. Given the nature of these conditions, timely management increases the probability of complete recovery. Rapid counteracting measures such as oral (or alternatively intravenous) rehydration or temporary renal replacement therapy may be key to restoring failed renal function. This emphasizes the need to develop tools that enable diagnosis to be made with minimal delay.

While clinical presentation and medical history are important indicators of the underlying condition, AKI-specific symptoms may be concealed, often within a myriad of nonspecific symptoms related to the underlying disease. Technological advances in the last decade have resulted in point-of-care testing (POCT) devices, which allow rapid assessments of not only renal parameters but also other factors ranging from metabolic and hormonal activities, to even specific cardiac parameters such as troponins and brain natriuretic peptides [3,4]. These devices are generally costly and require training, calibration and regular comparison to laboratory assessments. Given the low level of available resources in healthcare settings in developing countries, a great need exists for low-cost and simple techniques, which are easy to use and which reliably indicate rapid changes in the renal function.

We have conducted the current study and analyses as an advancement to our earlier studies, in which we have shown that saliva urea nitrogen (SUN) measured by a simple dipstick method reflects elevated blood urea nitrogen (BUN) in a cohort of patients with chronic kidney disease [5], and permits the determination of AKI disease severity at hospital admission [6]. In this study, we have carried the analyses another step forward by investigating (a) the agreement between SUN and BUN and (b) the diagnostic performance of SUN to discriminate more advanced AKI from less severe cases longitudinally over a time period of up to 8 days.

This international, multicenter, prospective cohort study enrolled patients diagnosed with AKI as per the AKI Network (AKIN) criteria [7] upon first presentation in the respective medical facility. Data were obtained by several investigators at Beth Israel Medical Center, New York, USA; Hans Dieter Schmidt Regional Hospital, Joinville, Santa Catarina, Brazil; and University Cajuru Hospital, Curitiba, Brazil. Institutional Review Boards of all involved institutions have approved the study and the merging of the databases in expedited review.

Unstimulated saliva was collected within 4 h of blood collection. Subjects were asked to refrain from drinking and eating for at least 15 min prior to saliva collection. Saliva was collected in a plastic cup and approximately 50 µl of saliva was used to moisten the colorimetric SUN dipstick (Integrated Biomedical Technology, Elkhart, Ind., USA). After 1 min, the color of the test pad was compared to 6 standardized color fields indicating SUN concentrations of 5-14 (color pad #1), 15-24 (#2), 25-34 (#3), 35-54 (#4), 55-74 (#5), and ≥75 (#6) mg/dl, respectively [5].

Parametric data are presented as mean ± standard deviation (SD) and non-parametric data are presented as median and interquartile range (IQR). The agreement between SUN and BUN was tested employing the analysis of variance with a post hoc Bonferroni correction for multiple testing. Differences between BUN and SUN (transformed to a continuous variable by choosing the midpoint for each range) were displayed as error bars (with 95% CIs) at different days and depicted as a modified Bland-Altman plot using BUN as the ‘gold-standard' [8,9]. Agreement between SUN and BUN over the entire period was also tested by the development of linear mixed effects models, which assumes different random intercepts for each subject and random slopes for each day-to-day period to account for variations between the days where measurements were conducted. Diagnostic performance of SUN and BUN to discriminate AKIN III vs. earlier stages was analyzed by predictive values (i.e. sensitivity and specificity after defining AKIN III as the binary outcome), and the area under the receiver operating characteristics curve at each of the 4 observation days following the approach defined by Calice-Silva et al. [6].

A two-sided p value <0.05 was considered statistically significant. Analyses were done in R 3.2.1 (codename ‘World-Famous Astronaut'; R Foundation for Statistical Computing, Vienna, Austria) additionally using the packages plyr, sandwich, nlme, multcomp, pROC and ggplot2 [10].

Forty patients were recruited; 3 of them had insufficient documentation and therefore 37 were studied in the primary analysis analyzing the agreement between SUN and BUN. Patients were followed over a variable length of time (median 4 (IQR 2-6.8) days), limited by a maximal observation period of 8 days. For the secondary analysis assessing the diagnostic performance of the SUN dipstick, we had to restrict the analysis to 31 subjects because of missing data. The characteristics of all patients are shown in table 1. It is of note that SUN, BUN and creatinine were substantially higher for those classified as AKIN III compared to the less severe AKIN stages I and II (table 1).

From an elevated baseline level for most patients, there was a clear and discernible trend toward a paralleling decrease in both parameters (fig. 1). A consistent decrease in serum creatinine can be seen in online supplemental figure 1 (for all online suppl. material, see www.karger.com/doi/10.1159/000445041). Table 2 shows a consistent underestimation of BUN by the SUN (ranging from 17.2 (day 4) to 25.7 (day 2)), which is also seen in figures 2 and 3. Pooling all data, regardless of the observation day, and analyzing BUN stratified by measured SUN categories showed significant differences between testpads 1 and 2 and testpads 3, 4, 5 and 6; and significant differences between testpads 3, 4 and 5 and testpad 6 (table 3). Table 3 shows a consistent underestimation of BUN by SUN, which is also seen in figure 2. Plotting the differences as a function of BUN and as a modified Bland-Altman plot showed consistent nonsignificant biases and proportional errors (fig. 4).

BUN stratified by measured SUN categories furthermore showed significant differences between testpad 1 and testpads 3, 4, 5 and 6 and significant differences between testpads 3, 4 and 5 and testpad 6. While these differences imply the ability to diagnose elevated SUN (and consequently BUN) with sufficient diagnostic accuracy, the lack of difference at lower levels also suggests a lower accuracy when SUN and BUN are low. One possible interpretation is that the dipstick appears to be a useful tool to diagnose very high SUN but not the lower levels (as often found in newly developing AKI), where it may misdiagnose and result in false-negative results. Currently, this is a limitation that may be addressed in future developments of the method.

Development of two linear mixed effects models, one with random intercept and another with random intercept (individual patient) and slope (individual observation day), with SUN as the dependent variable and BUN as fixed effects, was not significantly different in the likelihood ratio test. BUN was a significant fixed effect of SUN in both mixed effects models (0.52 mg/dl SUN per mg/dl BUN; p < 0.001 with an intercept of 3.7 and 3.2 mg/dl, respectively; p = 0.42).

Evaluating the diagnostic performance of BUN and SUN, respectively, showed both parameters, applying the criteria of Calice-Silva et al. [6], enabling the discrimination of AKIN III vs. AKIN I and II on all observation days. The area under the receiver operator characteristics curve decreased from 0.81 (95% CI 0.63-0.98) on observation day 1 to 0.57 (95% CI 0.63-0.98) on observation day 4 (fig. 5) for BUN. For SUN, the area under the receiver operator characteristics curve decreased from 0.85 (95% CI 0.71-0.98) on observation day 1 to 0.63 (95% CI 0.63-0.98) on observation day 4 (fig. 6). It is of note that the predictive value of serum creatinine did not change during the observation period (0.90 (95% CI 0.76-1.00) on observation day 1 to 0.91 (95% CI 0.76-1.00) on observation day 4 (online suppl. fig. 2)).

These analyses confirm our earlier reports on the agreement between SUN and BUN [5,6] and carry current knowledge on the diagnostic performance of SUN an additional step forward by analyzing the agreement and the diagnostic utility of SUN over a prolonged period of time (4 days) following first AKI diagnosis (as per the AKIN criteria [7]). We have found that the agreement between BUN and SUN was comparable on all observation days, generally following a declining trend in BUN and creatinine (fig. 1 and online suppl. fig. 1). While there is a consistent underestimation of BUN, SUN does, particularly at higher levels (>50 mg/dl), reliably reflect BUN and follows its changes over time (fig. 2, 3, 4). Table 3 shows a consistent underestimation of BUN by SUN, which decreases as BUN rises, which is also seen in figure 2. This is also seen in the systemic bias as shown in figure 4; however, for the purpose of discrimination of levels indicating the presence of AKI at AKIN III, where BUN is higher, the underestimation appears not to affect diagnostic performance when using SUN.

SUN in our data was consistent with the report of Calice-Silva et al. [6]; it was able to discriminate AKI at AKIN stage III vs. the less severe stages I and II, with a diagnostic performance comparable to BUN on every analyzed observation day (fig. 5 and 6). As to be expected, the diagnostic performance of both parameters decreased during the course of the study, likely because of the response to treatment; however, the performance of SUN was consistent with BUN on every observation day and discriminated with comparable accuracy (fig. 5 and 6).

Our data suggest that irrespective of the timing of diagnosis with regard to the onset of AKI, high SUN should immediately cause concern and trigger AKI-specific measures such as removal of nephrotoxins, assessment of fluid status, treatment of the underlying disease (e.g. infection), and possibly transfer to the next higher level of healthcare for more specific supportive measures, such as temporary dialysis. Furthermore, given the consistent agreement on all observation days, the dipstick appears to be a suitable tool to follow treatment progression. Differences of salivary flow rate in terms of age, gender, race and socioeconomic factors are a drawback of the dipstick method [11,12]. Given the considerably small sample size, interpretation must be made with caution; also, the fact that measurements were obtained from Brazil and the United States, and that 52% of the study participants were of black race lead to a result with some degree of external validity. Inter-observer variability was not investigated in the current analysis; however, based on our earlier results, these effects can be considered negligible [5]. In addition to well-known limitations of BUN as a marker of AKI, limitations of the SUN dipsticks such as the underestimation of BUN at levels below 50 mg/dl, with, the proportional error decreasing with the increase in BUN and the occurrence of false negative assessments are recognized. However, there are major strengths to this technique that need to be outlined. Currently, available POCT devices are expensive and, in addition, the test materials required for each measurement are expensive. In addition enzyme-containing cartridges need controlled cooled storage. Furthermore, the operating temperature maxima of many devices are exceeded by the ambient temperature in many developing countries. While in centralized healthcare facilities these devices are valuable, in remote areas with limited resources, lack of electricity, absence of a clean and ventilated/air-conditioned environment and lack of adequate training of involved personnel are obstacles to routine use. Considering the low cost of less than 1 US dollar (compared to around 7$ for a BUN measurement in the US) of the SUN dipstick, the remarkable ease of use and the low level of inter-observer variability, it is a particularly useful and promising tool for the diagnosis and follow-up of AKI especially in places where facilities are absent.

In view of the currently launched global efforts to eradicate mortality due to reversible AKI by the ‘0by25 Initiative' of the ISN, there is great interest in developing new diagnostic tools to diagnose AKI and disease severity. We believe that the SUN dipstick could be of a great utility in the diagnosis of AKI and the evaluation of disease severity, particularly in the most remote settings.

We acknowledge the generous provision of SUN dipstick free of charge by Integrated Biomedical Technology. The results presented in this paper have not been published previously in whole or part, except in the abstract form at the LI ERA-EDTA 2014 congress in Amsterdam, the Netherlands.

This study was partially funded by ISN - Research and Prevention Program - Grant. Saliva dipsticks for detection of urea nitrogen levels were provided free of charge by the manufacturer (Integrated Biomedical Technology, Elkhart, Ind., USA).

Dr. V. Calice-Silva was a fellow of the ISN and received scholarship from the Brazilian Government (CAPES) during part of the time when the study was conducted. Drs. J. Callegari, M. Carter, P. Kotanko and N.W. Levin hold stock in Fresenius Medical Care. All other authors have no financial interests to declare.

During this study, RPF received a scholarship from the Brazilian Council for Research Support (CNPq).

J.G.R. and V.C.-S. contributed equally to this work.