Lasers Med Sci DOI 10.1007/s10103-017-2288-5

ORIGINAL ARTICLE

Discrimination model applied to urinalysis of patients with diabetes and hypertension aiming at diagnosis of chronic kidney disease by Raman spectroscopy Elzo Everton de Souza Vieira 1,2 & Jeyse Aliana Martins Bispo 1,2 & Landulfo Silveira Jr 3,4 & Adriana Barrinha Fernandes 3,4

Received: 18 March 2016 / Accepted: 13 July 2017 # Springer-Verlag London Ltd. 2017

Abstract Higher blood pressure level and poor glycemic control in diabetic patients are considered progression factors that cause faster decline in kidney functions leading to kidney damage. The present study aimed to develop a quantification model of biomarkers creatinine, urea, and glucose by means of selected peaks of these compounds, measured by Raman spectroscopy, and to estimate the concentration of these analytes in the urine of normal subjects (G_N), diabetic patients with hypertension (G_WOL) patients with chronic renal failure doing dialysis (G_D). Raman peak intensities at 680 cm−1 (creatinine), 1004 cm−1 (urea), and 1128 cm−1 (glucose) from normal, diabetic, and hypertensive and doing dialysis patients, obtained with a dispersive 830 nm Raman spectrometer, were estimated through Origin software. Spectra of creatinine, urea, and glucose diluted in water were also obtained, and the same peaks were evaluated. A discrimination model based on Mahalanobis distance was developed. It was possible to determine the concentration of creatinine, urea, and glucose by means of the Raman peaks of the selected

* Adriana Barrinha Fernandes [email protected] 1

Faculdades Integradas do Tapajós – FIT, Rua Rosa Vermelha, 335, Aeroporto Velho, Santarém, Pará 68010-200, Brazil

2

Universidade Camilo Castelo Branco-UNICASTELO, Biomedical Engineering Institute, Parque Tecnológico de São José dos Campos, Estr. Dr. Altino Bondensan, 500, São José dos Campos, São Paulo 12247-016, Brazil

3

Biomedical Engineering Center, Universidade Anhembi Morumbi – UAM, Parque Tecnológico de São José dos Campos, Estr. Dr. Altino Bondensan, 500, São José dos Campos, São Paulo 12247-016, Brazil

4

Center of Innovation, Technology and Education – CITE, Parque Tecnológico de São José dos Campos, Estr. Dr. Altino Bondensan, 500, São José dos Campos, São Paulo 12247-016, Brazil

biomarkers in the urine of the groups G_N, G_WOL, and G_D (r = 0.9). It was shown that the groups G_WOL and G_D had lower creatinine and urea concentrations than the group G_N (p < 0.05). The classification model based on Mahalanobis distance applied to the concentrations of creatinine, urea, and glucose presented a correct classification of 89% for G_N, 86% for G_WOL, and 79% for G_D. It was possible to obtain quantitative information regarding important biomarkers in urine for the assessment of renal impairment in patients with diabetes and hypertension, and this information can be correlated with clinical criteria for the diagnosis of chronic kidney disease. Keywords Raman spectroscopy . Mahalanobis distance . Urinalysis . Diabetes . Hypertension . Chronic kidney disease

Introduction Chronic kidney disease (CKD) is considered a public health problem worldwide. There is an increasing incidence and prevalence of patients with kidney failure requiring replacement therapy, with high costs and poor results [1]. In Brazil, incidence and prevalence of end stage renal failure are increasing; prognosis is still bad, and costs of disease treatment are elevating [2]. The estimated total number of dialysis patients in Brazil in 2012 was 97,586. The number has increased gradually over the years, being 42,695 in the year 2000, 92,091 in 2010, and 91,314 in 2011. There was an annual increase of 3% compared to 2010. The prevalence rate of dialysis in 2012 was 503 patients per million of the population [2]. The major function of the kidneys is to remove waste products and excess fluid from the body. These waste products and excess fluids are removed through the urine. The production of urine involves highly complex steps of excretion and re-

Lasers Med Sci

absorption. This process is necessary to maintain a stable balance of body chemicals [3]. The process of progressive retention of metabolic nitrogen (uremia) and the failure of tubular function are among the main consequences of renal failure. The failure of tubular function produces an early inability to concentrate urine, fluid retention and abnormalities in biochemical homeostasis (including the reaction of water and salt, compensated metabolic acidosis, and other electrolyte imbalances), and hypertension [4]. Chronic kidney disease (CKC) is characterized by the reduced ability of the kidney to carry out these functions in the long-term. The absence of symptoms in patients who are in the early stages of CKD requires clinicians to always maintain an appropriate level of suspicion, especially in those patients with medical problems such as cardiovascular disease, diabetes, and hypertension. Higher blood pressure level and poor glycemic control in diabetes are considered progression factor that should cause worsening kidney damage and faster decline in kidney function after initiation of kidney damage [5]. In individuals with diabetes, hyperglycemia can be severe. Glucosuria can also occur when the renal tubular transporter system for glucose becomes saturated. This happens when the glucose concentration of plasma exceeds roughly 180 mg/dL in an individual with normal renal function and urine output. As hepatic glucose overproduction continues, the plasma glucose concentration reaches a plateau around 300 to 500 mg/dL (17–28 mmol/L) [6]. Creatinine plays important roles in many biological processes [7]. Plasma creatinine is inversely related to glomerular filtration rate (GFR—the volume of plasma filtered by glomerulus per unit of time) and, although an imperfect measure, is commonly used to assess renal filtration function [8]. Creatinine clearance, a measure of the amount of creatinine eliminated from the blood by kidneys, is also used to gauge renal function [9]. Plasma and urinary quantification of creatinine concentration is used to determine sufficiency of kidney function and the severity of kidney damage and is employed to monitor the progression of kidney disease [9]. The methods most frequently used to measure creatinine are based on the Jaffe reaction first described in 1886 [10]. The procedure initially described was modified, allowing direct measurements in serum, plasma, and urine [6]. Urea production occurs almost exclusively in the liver due to protein metabolism, and it is carried in the blood to the kidney, where it is filtered from the plasma by the glomerulus. Most of the urea is excreted in the urine, although some urea is reabsorbed by passive diffusion during passage of the filtrate through the renal tubules. The amount reabsorbed depends on urine flow rate and extent of hydration; only small quantities of urea are excreted through the skin and gastrointestinal tract. The plasma urea concentration is determined by renal function and perfusion, the protein content of the diet, and the rate of

protein catabolism [6]. Quantification of urea concentration is important in evaluating renal function (diagnosing renal disease), verifying adequacy of dialysis, determining nitrogen balance, and assessing hydration status [11]. The early treatment of acute renal failure can be correlated with a better prognosis of renal diseases, and the identification of urine biomarkers for early diagnosis would improve the efficacy of a therapeutic strategy [12]. Thus, it becomes imperative to determine techniques for a rapid quantification of urine biomarkers that can stratify correctly the extent of renal damage that each patient has suffered and the risk of developing CKD. Raman spectroscopy is an optical technique which uses laser light to assess the vibrational energy levels of the molecules through inelastic scattering of incident radiation by the polarized molecules [13]. The Raman spectrum provides biochemical and structural information of organic and inorganic compounds in biological tissues and fluids, allowing their rapid identification of the chemical structure of a material, with little or no need for sample preparation or reagents [13]. Raman spectroscopy has been employed to estimate the concentration of different biochemicals related to diseases in human serum [14, 15] and urine [16, 17]. Raman spectroscopy for urine analysis offers several advantages such as urine analysis without dilution, no need for reagents, a small volume of sample is needed, and multiple components can be evaluated at a time [13, 16, 18]. Raman analysis can benefit of spot urine and/or first morning urine collections, as these single samples present comparable results to the 24-h collection for clinical end-points and outpatients [19, 20]. Premasiri et al. (2001) employed conventional Raman spectroscopy and SERS (surface-enhanced Raman spectroscopy) to obtain the spectra of the components creatinine, urea, and uric acid present in human urine, and the authors showed that SERS was adequate to quantitatively evaluate the creatinine present in urine [18]. McMurdy and Berger (2003) reported the first use of Raman spectroscopy to estimate the concentration of creatinine in unaltered urine samples from a multipatient population, obtaining prediction error as low as 4.9 mg/dL [16]. Park et al. (2007) used low resolution Raman spectrometer to detect diluted urine samples with minute amounts of glucose simulating abnormal concentrations of this metabolite aiming diabetes diagnosis [21]. Wang et al. (2010) prepared nanostructured surfaceenhanced Raman spectroscopy SERS substrates for the quantitative analysis of creatinine concentration in urine and obtained highly sensitive creatinine concentration of 0.5 g/mL and reproducible 5% variation detection, and the authors showed good agreement with the standard enzymatic method [22]. Li et al. (2015) reported a novel reagent- and separation-free method for urine creatinine concentration measurement using stamping surface-enhanced Raman scattering (S-SERS) technique with nanoporous gold disk (NPGD) plasmonic substrates, being the limit of detection of 0.15 μg/dL and 0.68 mg/dL in water and urine, respectively [23]. Saatkamp et al. (2016) employed a

Lasers Med Sci

spectral model to estimate the concentrations of urea and creatinine in urine by means of dispersive Raman spectroscopy, and the partial least squares (PLS) models presented cross-validation errors of 312 and 25.2 mg/dL for urea and creatinine, respectively, that could be used to diagnose kidney disease [17]. The present study aimed to develop a quantification model to assess the biomarkers creatinine, urea, and glucose in first morning urine samples of groups of normal subjects, patients with diabetes mellitus and hypertension, and patients with chronic renal failure doing dialysis by means of Raman spectroscopy. The concentrations of these compounds were used to develop a diagnostic model based on discriminant analysis employing Mahalanobis distance to classify urine samples in one of the three groups aiming to use the chemical information provided by Raman spectra for prediction of renal injury. Participants and methods This study was approved by Research Ethics Committee from UNICASTELO (protocol nos. 64116 and 8926), according to Guidelines and regulating norms for research involving human subjects (resolution no. 196/96). The present work used the spectra of urine from the study of Bispo et al. [24]. Briefly, a total of 72 patients (42 women and 30 men) were enrolled, which were divided as follows: 18 normoglycemic and normotensive volunteers (G_N), 35 diabetic and hypertensive patients (G_WOL), and 19 diabetic and hypertensive patients who had renal failure and are being submitted to blood dialysis (G_D). Figure 1 shows the illustrative diagram to demonstrate the criteria of the inclusion and exclusion of participants in the present study. First morning urine samples were collected in fasting, bottled and stored in −80 °C freezer until spectral analysis. The samples of normoglycemic and normotensive volunteers (G_N) and diabetic and hypertensive patients (G_WOL) were obtained in the Hyperdia Group organized by the Municipal Department of Health (Santarém/Pará) in the Health Unit of the Faculdades Integradas Tapajós (FIT), whereas urine samples of diabetic and hypertensive patients who have renal failure and are being submitted to blood dialysis (G_D) were collected before the hemodialysis session (predialysis) in the Municipal Hospital of Santarém and the Lower Amazon Regional Hospital (Santarém-Pará). Prior Raman spectroscopy analysis, urine samples were unfrozen to reach room temperature and placed in an aluminum holder (Bhome^ made, 10 mm height, 50 mm width, 100 mm length) with six vessels of about 80 μL. Spectra were taken in the vessel by means of a Raman probe connected to a dispersive Raman spectrometer. The Raman spectrometer (model P-1, Lambda Solutions, Waltham, MA, USA) is composed of a diode laser (830 nm) coupled to a probe (model Vector Probe, Lambda Solutions, Waltham, MA, USA) that is used to illuminate sample and

collect the scattered light, as described in Bispo et al. [24]. The probe is connected to a spectrograph and CCD camera, which collects high-resolution Raman spectrum from the sample in the fingerprint region (400 to 1800 cm−1). The laser power was adjusted to 300 mW, and the integration time to collect the Raman signal was set to 20 s. Triplicate spectra were obtained from each sample, which were averaged after preprocessing. In order to quantify the constituents of urine, the intensities and areas of the peaks of creatinine (680 cm−1), urea (1004 cm−1), and glucose (1128 cm−1) were obtained for each spectrum using the Gaussian curve fitting of the software Origin (version 6, Microcal Software, Inc., Northampton, MA, USA). We have obtained calibration curves for quantification of urine components: creatinine (ref. C4255, Sigma-Aldrich Brazil, São Paulo, Brazil), urea (ref. 161-0730, Bio-Rad Laboratory, Rio de Janeiro, Brazil), and glucose (ref. 49163, Sigma-Aldrich Brazil, São Paulo, Brazil). Calibration curves were derived from dilutions of creatinine (20, 40, 80, 160, 240, and 320 mg/ L), urea (25, 50, 125, 250, and 500 mg/mL), and glucose (250, 500, 1000, and 2000 mg/mL) in distilled water, being the concentrations within the physiological range. The peak intensity versus concentration (mg/mL) of each analyte was plotted, and the equations curves were determined. Then, the concentration of each analyte in urine was obtained using the equations and the peak intensities. The average concentration of each analyte in each group was plotted. The analysis of variance (ANOVA) with 5% significance (p < 0.05) was applied to determine whether the groups had significant differences in creatinine, urea, and glucose. In order to discriminate the urine samples in groups normal (G_N), diabetic and hypertensive patients without kidney injury (G_WOL) and diabetic and hypertensive renal injury performing hemodialysis (G_D), it developed a discrimination model based on Mahalanobis distance using the estimated concentrations of the analytes urea, creatinine, and glucose. The Mahalanobis distance (also called m-distance) is very useful to find the similarity of a set of values from one group of samples compared to another group in discriminant analysis. The Mahalanobis distance (D2) is calculated as follow [25]: D2i ¼ ðx−μi ÞT V −1 ðx−μi Þ where i denotes the known sample group number, x is the vector of the sample parameter, μ is the mean vector for the specific group, and V is the covariance matrix of the group. Since Mahalanobis distance takes into account the covariance matrix of the data, it gives a statistical of how well the sample matches the grouped data and can be used to classify the sample into well-defined classes.

Lasers Med Sci Fig. 1 Illustrative diagram to demonstrate the criteria of inclusion and exclusion of participants

Assessed for eligibility (n=100) Excluded (n=28) Not meeng unclusion criteria (n=28) Refused to parcipe (n=0) Others reason (n=0)

Allocated to G_N= Group Normal (n=18)

Allocated to G_WOL= Group Without Lesion (n=35)

-Aged 50-80 years -Both gender - Normoglycemic and normotensive volunteers - Fasng for at least two hours.

-Aged 50-80 years - Both gender - Diabec and hypertensive paents - Taking insuline regularly - Controlling blood pressure - Fasng for at least two hours.

Discrimination using Mahalanobis distance was obtained using the function Bclassify.m^ [26] in Matlab software (version 7.4.0, The Mathworks, Natick, MA, USA) where the input variables Bsample^ and Btraining^ are the concentrations of biochemical elements of the sample belonging to each group (G_N, G_WOL, and G_D), and the Bgroup^ is a column of numbers representing the three classes of the clinical conditions. We tested discrimination using the Raman concentrations of creatinine and urea, and subsequently, it was included glucose.

Results The mean spectra of urine from the groups G_N, G_WOL, and G_D are presented in Fig. 2a. Figure 2b presents the Raman spectra of water diluted creatinine, urea, and glucose. In the spectra of urines, it was seen peaks related to urea, creatinine, glucose, and other biochemical. The intensity of these main peaks was used to estimate the concentrations of the analytes of interest. Figure 3 presents the calibration curves based on the intensity of the peaks creatinine, urea, and glucose (680, 1004, and 1128 cm−1, respectively) versus the concentrations of these compounds diluted in distilled water. Linearity and high correlation were observed in the concentrations of the analytes (R2 > 0.98 for all biochemical). The concentrations of urea (mg/dL) obtained for the groups G-N and G_WOL were 277.6 ± 75.9 and 165.3 ± 56.8, respectively, and concentrations of creatinine (mg/dL) obtained for the groups G-N and G_WOL, 124.4 ± 48.5 and 55.3 mg/dL ± 38.9, respectively, through colorimetric method. But for the group dialysis, the volume of urine for most samples was not sufficient to perform the biochemical analysis.

Allocated to G_D= Group Dialysis (n=19)

-Aged 50-80 years - Both gender -Diabec and hypertensive paents -Renal failure and are doing blood dialysis - Taking insuline regularly - Controlling blood pressure - Fasng for at least two hours.

Figure 4a presents the mean intensities of the Raman peaks of creatinine, urea, and glucose in the different groups, and Fig. 4b presents the average concentrations of these analytes obtained by the respective equations of Fig. 3. There was a statistically significant difference in urea and creatinine concentrations in the different groups compared to the group G_N (ANOVA, p < 0.05, Fig. 3b), being these values decreased for the clinical groups G_WOL and G_D. Despite not significant due to the large standard deviation, patients in the groups G_WOL and G_D have a greater glucose concentration in urine than the G_N group. In order to discriminate between the groups: normal subjects (G_N), patients with diabetes and hypertension (G_WOL) and patients with CKD performing dialysis (G_D), we developed discrimination models based on Mahalanobis distance using the concentrations of the biochemicals creatinine and urea, and then including glucose. Figure 5a shows discrimination according to the concentrations of creatinine and urea, and Fig. 5b shows the discrimination according to the concentration of creatinine, urea, and glucose. According to the results presented in Table 1, the percentage of correct classification was 89% for G_N, 86% for G_WOL, and 79% for G_D, considering the use of creatinine, urea, and glucose as discriminant variables.

Discussion Previous studies demonstrated that Raman spectroscopy of urine is a promising tool for identification and quantification analysis of potential biomarkers for kidney disease such as creatinine [16, 18, 22] and urea [17, 24]. Urine is composed basically of urea, creatinine, and many other metabolites but

a

G_N (normal) G_WOL (non-dialysis)

680 creat

1128 gluc

G_D (dialysis)

Intensity (arb. un.)

Fig. 2 a Mean Raman spectra of the urine samples of the groups normal (G_N), without renal impairment (G_WOL), and performing dialysis (G_D). Marked peaks are due to creatinine (680 cm−1), urea (1004 cm−1), and glucose (1128 cm−1). b Raman spectra of water-diluted creatinine, urea, and glucose

1004 urea

Lasers Med Sci

400

600

800

1000

1200

1400

1600

1800

-1

680

Raman shift (cm )

1128

1004

Intensity (arb. un.)

b

Creatinine

400

600

800

1000

1200

Urea

1400

Glucose

1600

1800

-1

Raman shift (cm )

in lower concentrations (less than mg/dL). The spectral signature of urine has been presented by Premasiri [16] and Bispo [24], and these compounds present peaks in different positions (no overlap). In the peak positions of urea, creatinine, and glucose, no peak from other compound has been described so far. In a previous study, Bispo et al. [24] aimed to identify potential biomarkers in urine of diabetic and hypertensive patients who presented or not complications related to these disorders, and the groups were discriminated through principal components analysis (PCA) and discriminant analysis based on quadratic distance. The discriminating model separated the urine samples in four groups according to the clinical evaluation with high accuracy (70% overall accuracy). This study demonstrated that Raman technique is a fast and reliable method for qualitative assessment of urines from patients with diabetes and hypertension and may be useful in the diagnosis of complications associated to these diseases. In the present study, the concentrations of the biomarkers creatinine, urea, and glucose were estimated in the Raman spectra of urine using the peak intensities of each analyte. The groups G_WOL and G_D presented significantly lower creatinine and urea concentrations than the group G_N. Lower creatinine and urea values suggest decreased renal function

due to diabetes and hypertension in the group G_WOL and G_D. These data are consistent with the clinical finding, since the G_D group is composed of individuals with CKD doing dialysis. However, patients G_WOL group showed no significant difference with respect to G_D group in the concentration of creatinine, which is a worrying fact, since some patients in the group G_WOL present clinical complications due to diabetes and hypertension, which may indicate patients near the condition of kidney failure, despite the creatinine biomarker is not specific to monitoring of CKD [6]. Patients of groups G_WOL and G_D have a greater glucose concentration in urine than the G_N group, not significant probably due to the large standard deviation. Glucosuria reported by patients of G_WOL and G_D groups, although present since they are patients with diagnosed diabetes mellitus, is indeed worrisome, because it indicates that the therapy for diabetes is not being effective, since the presence of glucose in urine reflects the inability in the tubules to retain this compound due to the specific glucose tubular injury. In this case, the renal threshold (capacity of the tubules to reabsorb glucose) has been reached, reflecting the occurrence of renal glucosuria, which may be present in association with other disorders in the tubular function [6].

Lasers Med Sci 0.8

680 cm-1 intensity (a.u.)

Fig. 3 Plotting the peak intensities: 680 cm−1 (creatinine), 1004 cm−1 (urea), and 1128 cm−1 (glucose) versus the concentration of these compounds diluted in deionized water. Equations and correlation coefficients are shown in the graphs

y = 0.00180x + 0.0725 2 R = 0.989

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0

50

100

150

200

250

300

350

1004 cm-1 intensity (a.u.)

Creatinine concentration (mg/dL) 10 y = 0.0184x + 0.110 2

R = 0.982

5

0 0

100

200

300

400

500

600

Urea concentration (mg/dL)

1128 cm-1 intensity

2.5 y = 0.000944x + 0.0511 2 R = 0.999

2.0 1.5 1.0 0.5 0.0 0

500

1000

1500

2000

2500

Glucose concentration (mg/dL)

Discrimination of hemoglobin S (sickle cell disease) from hemoglobin A (normal) in human blood using Mahalanobis distance has been presented by Bueno Filho et al. [27], with 100% success of classification of hemoglobin S. In the present work, we used discriminant analysis that employs the Bclassify.m^ function of Matlab software, being the discriminator of the Mahalanobis distance. Therefore, the output is the class number that is obtained depending on the Mahalanobis distance of the particular point (represented by the three variables—the concentrations) to the group’s mean point. The discriminant model applied to the concentrations of creatinine, urea, and glucose grouped the samples in the clinical groups with high discrimination using the three biomarkers (overall diagnostic accuracy

of 85%). Considering the classification errors of the G_N group, two patients were classified in the G_WOL. The group G_WOL presented two patients classified in the group G_N and three patients in the group G_D. Four patients of the group G_D were classified as group G_WOL. Indeed, it was initially proposed to discrimination between the three clinical groups using creatinine and urea concentrations (Fig. 4a); however, the presence of glucose in the Raman spectra of urine from patients in the group G_WOL suggested including the glucose as an important biomarker for these patients that were being wrongly classified as group G_N. This suggests that glucose is an important parameter for discrimination of patients with alterations in the biochemical profile of

a

Creatinine conc. (mg/dL)

0.5 0.4 0.3 0.2 0.1 0.0

G_N

200 180 160 140 120 100 80 60 40 20 0

G_D 400

7.0

350

6.0 5.0 4.0 3.0 2.0 1.0

b a,b

a b

G_N

8.0

G_WOL

G_D

a,b

300

a,c

250 200 150

b,c

100 50 0

0.0

G_N

G_WOL

G_D

1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 -0.1 -0.2 -0.3

G_N

G_WOL

G_D

G_N

G_WOL

G_D

1000

Glucose conc. (mg/dL)

Glucose peak intens. (a.u.)

G_WOL

Urea conc. (mg/dL)

Urea peak intens. (a.u.)

Fig. 4 a Intensity mean and standard deviation of creatinine peaks (680 cm−1), urea (1004 cm−1), and glucose (1128 cm−1) in the different groups and concentration of these compounds in the samples obtained from the equations. b Concentrations of elements calculated by the biochemical quantitative model based on the intensities of the Raman peak. Superscript letters a, b, and c indicate statistically significant differences (ANOVA, p < 0.05) between groups

Creatinine peak intens. (a.u.)

Lasers Med Sci

800 600 400 200 0 -200 -400

G_N

G_WOL

urine, since it may indicate renal injury. Thus, the inclusion of glucose in the model was considered an important parameter to reclassify patients, since some patients in the group G-WOL were classified as normal using only the concentrations of creatinine and urea. Poor glycemic control in diabetic patients and high blood pressure levels are considered progression factors that should decline kidney function after initiation of kidney damage [5]. This study was performed in a group of patients with diabetes and hypertension. The reduced urea and creatinine is an indicative of renal lesion taking place; the presence of glucose is an indicative of poor diabetes control. Therefore, the outcome of the Raman analysis indicates the need for more specialized diagnosis of the clinical status of a diabetic and hypertensive subject: if urine falls into the group G-WOL, the patient needs to be submitted to an analysis of the kidney function and to perform a better control of diabetes and hypertension;

G_D

if it falls into the group G-D, the kidney function needs to be urgently evaluated. It is noteworthy that there are clinical and biochemical parameters that are used in order to perform a clinical classification of these patients, and these parameters are important on the primary care (prevention). However, it is observed in clinical practice that often patients are asymptomatic, and therefore do not seek specialized medical care, and consequently do not receive adequate treatment in order to prevent the progression of renal disease. In addition, methods of clinical and laboratory diagnosis employed currently have difficulties in the early diagnosis of kidney injury, which may be the case for G_WOL group of patients who might be on the verge of developing renal injury and which the spectral model indicted to belong to the group G_D. Therefore, the development of optical techniques for urinalysis may become a tool for early renal disease diagnosis and control for population screening.

Lasers Med Sci

a G_N G_WOL G_D

Classification by Mahalanobis creatinine and urea conc.

Fig. 5 Results of the discrimination in the normal groups (G_N), without renal impairment group (G_WOL), and performing dialysis group (G_D) using the discrimination model based on Mahalanobis distance applied to the concentrations. a Creatinine and urea to discriminate all the groups. b Glucose to better discriminate the cases of G-WOL group that were classified as G_N group

G_N

G_WOL

G_D

b G_N G_WOL G_D

Classification by Mahalanobis glucose concentration

Groups

G_N

G_WOL

G_D

Groups

Conclusion In this work, we present the possibility of obtaining the concentrations of creatinine, urea, and glucose in first morning urine samples using the peak intensity of these biochemicals in the Raman spectra of urine of patients with and without renal impairment. It was possible to develop a model for discrimination of patients with diabetes and hypertension and patients doing dialysis against control groups, using the concentrations of creatinine, urea, and glucose, using the Mahalanobis distance as discrimination parameter. Raman technique has been shown to be a

Table 1 Percentage of success of classification of the groups according to the discrimination model based on the urinary concentrations of creatinine, urea, and glucose Discrimination by Bclassify.m^ and Mahalanobis distance Groups based on clinical criteria G_N (n = 18) G_WOL (n = 35) G_D (n = 19)

G_N

G_WOL

G_D

Successes (%)

16 2 –

2 30 4

– 3 15

89 86 79

promising tool for quantitative analysis of potential biomarkers of renal function such as creatinine, urea, and glucose in recent urine. It has been shown that it is possible to obtain quantitative information about the important biomarkers for the assessment of renal impairment in patients with diabetes mellitus and hypertension, and these can be correlated with clinical criteria for the diagnosis of CKD and may be used to predict its occurrence.

Acknowledgments L. Silveira Jr. thanks the FAPESP (São Paulo Research Foundation) for the partial financial support (process no. 2009/01788-5) and CNPq (National Council for Scientific and Technological Development) for the Productivity Fellowship (process no. 305680/2014-5). E. E. S. Vieira and J. A. M. Bispo thank the Regional Public Hospital of Western Pará and Santarém Municipal Hospital for authorizing this research in the Nephrology Clinic.

Compliance with ethical standards Conflict of interest The authors declare that they have no conflict of interest. Role of funding sources The project has been supported in part by a public funding agency (São Paulo Research Foundation—FAPESP), Process FAPESP no. 2009/01788-5, that allowed to purchase the Raman spectrometer.

Lasers Med Sci Ethical approval This study was approved by Research Ethics Committee from UNICASTELO (protocols nos. 64116 and 8926).

14.

Informed consent All volunteers agreed and signed an informed consent form to participate this study.

15.

References

16.

1.

2.

3. 4. 5.

6.

7. 8. 9.

10.

11.

12.

13.

Eknoyan G, Lameire N, Barsoum R, Eckardt KU, Levin A, Locatelli F et al (2004) The burden of kidney disease: improving global outcomes. Kidney Int 66:1310–1314 Sesso RCC, Lopes AA, Thomé FS, Lugon JR, Santos DR (2011) Chronic dialysis in Brazil—report of the Brazilian dialysis census. J Bras Nefrol 34:272–277 Hall JE (2011) Guyton and hall textbook of medical physiology, 12th edn. Elsevier Saunders, Philadelphia Kumar V, Abbas AK, Aster JC (2014) Robbins & Cotran pathologic basis of disease, 9th edn. Elsevier Saunders, Philadelphia National Kidney Foundation (2003) K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification and stratification. Am J Kidney Dis 39:1–266 Bishop ML, Fody EP, Shoeff LE (2013) Clinical chemistry: principles, techniques, and correlations, 7th edn. Lippincott Williams & Wilkins, Philadelphia Nelson DL, Cox MM (2008) Lehninger principles of biochemistry, 5th edn. W.H. Freeman, New York Apps DK, Cohen BB, Steel CM (1992) Biochemistry: a concise text for medical students, 5th edn. Bailliere Tindall, London Assessment of kidney function serum creatinine, BUN, and GFR, Retrieved Spring, 2008. Available at http://www.uptodate.com. Accessed 04 Dec 2014 Jaffe M (1986) Uber den niederschlag welchen pikrinsaure in normalen harn erzeugt und uber eine neue reaktion des kreatinins. Hoppe Seyler's Z Physiol Chem 10:391–400 Oh MS (2011) Evaluation of renal function, water, electrolytes and acid-base balance. In: McPerson RA, Pincus MR (eds) Henry's clinical diagnosis and management by laboratory methods, 22nd edn. Saunders Elsevier, Philadelphia Bastos MG, Kirsztajn GM (2011) Chronic kidney disease: importance of early diagnosis, immediate referral and structured interdisciplinary approach to improve outcomes in patients not yet on dialysis. J Bras Nefrol 22:93–108 Hanlon EB, Manoharan R, Koo TW, Shafer KE, Motz JT, Fitzmaurice M et al (2000) Prospects for in vivo Raman spectroscopy. Phys Med Biol 45:R1–R59

17.

18. 19.

20.

21.

22.

23.

24.

25.

26.

27.

De Almeida ML, Saatkamp CJ, Fernandes AB, Pinheiro AL, Silveira L (2016) Estimating the concentration of urea and creatinine in the human serum of normal and dialysis patients through Raman spectroscopy. Lasers Med Sci 31:1415–1423 Silveira L, Borges RC, Navarro RS, Giana HE, Zângaro RA, Pacheco MT, Fernandes AB (2017) Quantifying glucose and lipid components in human serum by Raman spectroscopy and multivariate statistics. Lasers Med Sci (in press) McMurdy JW, Berger AJ (2003) Raman spectroscopy-based creatinine measurement in urine samples from a multipatient population. Appl Spectrosc 57:522–525 Saatkamp CJ, de Almeida ML, Bispo JA, Pinheiro AL, Fernandes AB, Silveira L (2016) Quantifying creatinine and urea in human urine through Raman spectroscopy aiming at diagnosis of kidney disease. J Biomed Opt 21:37001 Premasiri WR, Clarke RH, Womble ME (2001) Urine analysis by laser Raman spectroscopy. Lasers Surg Med 28:330–334 Teo BW, Loh PT, Wong WK, Ho PJ, Choi KP, Toh QC, Xu H, Saw S, Lau T, Sethi S, Lee EJ (2015) Spot urine estimations are equivalent to 24-hour urine assessments of urine protein excretion for predicting clinical outcomes. Int J Nephrol 2015:156484 Xin G, Wang M, Jiao LL, Xu GB, Wang HY (2004) Protein-tocreatinine ratio in spot urine samples as a predictor of quantitation of proteinuria. Clin Chim Acta 350:35–39 Park C, Kim K, Choi J, Park K (2007) Classification of glucose concentration in diluted urine using the low-resolution Raman spectroscopy and kernel optimization methods. Physiol Meas 28:583–593 Wang H, Malvadkar N, Koytek S, Bylander J, Reeves WB, Demirel MC (2010) Quantitative analysis of creatinine in urine by metalized nanostructured parylene. J Biomed Opt 15:027004 Li M, Du Y, Zhao F, Zeng J, Mohan C, Shih WC (2015) Reagentand separation-free measurements of urine creatinine concentration using stamping surface enhanced Raman scattering (S-SERS). Biomed Opt Express 6:849–858 Bispo JA, Vieira EES, Silveira L, Fernandes AB (2013) Correlating the amount of urea, creatinine, and glucose in urine from patients with diabetes mellitus and hypertension with the risk of developing renal lesions by means of Raman spectroscopy and principal component analysis. J Biomed Opt 18:1–8 Ciaccio EJ, Dunn SM, Akay M (1994) Biosignal patternrecognition and interpretation systems. IEEE Eng Med Biol 13: 129–135 The MathWorks Inc (2016) Documentation. Classify. The Mathworks Inc. https://www.mathworks.com/help/stats/classify. html. Accessed 20 Oct 2016 Filho AC, Silveira L, Yanai AL, Fernandes AB (2015) Raman spectroscopy for a rapid diagnosis of sickle cell disease in human blood samples: a preliminary study. Lasers Med Sci 30:247–253