Evaluation of Criteria for the Diagnosis of Balkan Endemic Nephropathy

Abstract

Background. The diagnosis of Balkan endemic nephropathy (BEN) is often made using Danilovic's criteria. The aim of this study was to determine the prevalence, sensitivity, specificity, and predictive value of Danilovic's criteria and several additional indices. Methods. The study included 19 BEN patients, 23 BEN-suspected patients, 34 patients with other kidney diseases, and 23 healthy controls. The sensitivity and specificity of Danilovic's criteria was calculated, and these criteria, in addition to age, sex, blood pressure, creatinine clearance, glucosuria, urine osmolality, alkaline phosphatase, alpha 1-microglobulin, fractional sodium excretion, tubular phosphate reabsorption, kidney length, and volume, were combined in a logistic regression. Results. All examined persons were from a BEN-affected village (criterion 1), and all BEN, BEN-suspected patients, and 12/23 healthy controls were from BEN families (criterion 2). None of the remaining Danilovic's criteria was found in the healthy controls. The prevalence of proteinuria, low specific gravity, and anemia (criteria 3–5) differed insignificantly among the patient groups. Azotemia and shrunken kidney (criteria 6 and 7) were significantly more frequent in BEN than in other patients. Only proteinuria showed high sensitivity and specificity in differentiating BEN and BEN-suspected patients from healthy persons, but no criteria differentiated BEN or BEN-suspected from other kidney diseases. Proteinuria is a significant predictor of both BEN and BEN-suspected vs. healthy persons, and alpha 1-microglobulinuria is a significant predictor of BEN vs. other kidney diseases. Conclusion. Danilovic's criteria enabled a diagnosis of BEN only in chronic renal failure and differential diagnosis between BEN and healthy persons but not between BEN and other kidney diseases. Out of the examined indices of proximal tubular disorders, only alpha 1-microglobulinuria significantly discriminated BEN from other kidney diseases.

Keywords:

• Balkan endemic nephropathy
• diagnostic criteria
• sensitivity
• specificity
• predictive value

INTRODUCTION

The need to define methods and criteria for the diagnosis of Balkan endemic nephropathy (BEN) was already recognized during initial studies on the disease. However, after five decades of investigation, the diagnosis of BEN remains an unsolved problem. In 1979, Danilović proposed some criteria for the diagnosis of BEN:

1. farmers living in the endangered villages,

2. a familial history positive for BEN,

3. mild proteinuria,

4. low specific gravity of the urine,

5. anemia,

6. retention of nitrogen compounds in the blood (urea > 50 mg/dL, creatinine > 1.5 mg/dL), and

7. symmetrically shrunken kidneys.[1]

These or very similarly defined criteria were frequently used in studies on BEN.[2–5] Using them, Danilović also suggested classifying patients into the following groups:

1. patients with intermittent proteinuria who fulfill at least the first three criteria,

2. BEN-suspected patients who, in addition to fulfilling the first three criteria, fulfill at least one of the remaining criteria,

3. BEN patients who fulfill at least five criteria, and

4. BEN-decompensated patients who fulfill at least five criteria and have urea values >150 mg/dL and manifested signs of uremia[1]

This or a very similar classification system has been used by other authors as well.[4],[5]

Although it has been stressed that the criteria of Danilović are not sufficiently specific, there is no study on their specificity and sensitivity. Moreover, attempts to uncover more specific and sensitive diagnostic indices, especially for a diagnosis of the early phase of the disease, have not been successful.[4],[6],[7] Therefore, a diagnosis of BEN is still made by ruling out other kidney diseases and revealing a sufficient number of existing criteria.

The aim of the present study was to determine the prevalence, sensitivity, and specificity of the criteria for BEN diagnosis proposed by Danilović and examine the predictive value of these criteria and a number of additional indices relating to kidney size and function.

SUBJECTS AND METHODS

The study included 99 persons selected in a cross-sectional study of a BEN-affected village (Vreoci, Kolubara region, Serbia) that involved 2009 (81.6%) out of 2462 inhabitants over the age of 18 years. During three phases of the study comprising epidemiological, clinical, and laboratory examinations described elsewhere,[8] the following groups of patients were detected:

• group 1: 19 BEN patients,

• group 2: 23 BEN-suspected patients,

• group 3: 34 patients with other kidney diseases.

In addition, a fourth group with 23 healthy controls was included in the study. The diagnosis of BEN and suspected BEN was established according to the criteria of Danilovic,[1] and other kidney diseases were ruled out. Kidney diseases other than BEN were diagnosed using routine laboratory, x-ray and sonographic methods, including kidney biopsy in patients with glomerulonephritis. The underlying kidney disease in group 3 was glomerulonephritis in 10 patients, pyelonephritis in 6 patients, diabetic nephropathy in 6 patients, reflux nephropathy in 4 patients, and nephrosclerosis in 8 patients. Group 4 comprised 23 healthy persons, randomly selected from a population of healthy persons detected in the cross-sectional study of the village of Vreoci. All of them had a negative medical history for kidney disease, hypertension, and diabetes, and no pathologic finding was detected by objective, laboratory, or sonographic examination.

All subjects included in the study gave their medical history, completed an epidemiological questionnaire, and were objectively examined. Body mass index (BMI) was calculated according to the following formula:

Body surface area (BSA) was estimated according to weight and height by using a nomogram based on the formula of DuBois.[9] Arterial hypertension was diagnosed when systolic blood pressure was ≥140 mmHg and/or diastolic pressure was ≥90 mmHg, or if antihypertensive treatment was prescribed.

The Ethics Committee of the Clinical Center of Serbia evaluated and approved this study, and both patients and healthy controls gave their informed consent.

Laboratory analyses included peripheral blood cell count and serum and urine levels of urea, sodium, and phosphate measured with a commercially available kit (Beckman, Germany) using Synchron CX Beckman (Brea, California, USA). Urine glucose was measured as a “spot test” with a dipstick containing a color-sensitive pad. Urine alpha 1-microglobulin was measured by immunoturbidimetric assay (Turbitex α₁-microglobulin; Roche/Hitachi 902; normal value <1.5 mg/mmol creatinine) and alkaline phosphatase (AP) by a colorimetric method (normal value < 0.37 U/mmol creatinine). Morning urine specimens were used for all of the above mentioned analyses. Serum creatinine was determined on a Beckman Creatinine Analyzer II (Brea, California, USA) with the modified Jaffe rate method. Azotemia was defined by Danilovic as serum urea or creatinine level above the upper normal limit.

In the sample of urine collected over 24 hours, protein was measured by a biuret method and creatinine by the Jaffe method, and 24h urinary creatinine clearance was calculated. Proteinuria above 200 mg/day was considered pathologic. Fractional sodium excretion (FENa) was calculated using the following formula:

Percentage tubular phosphate reabsorption (TRP) was calculated according to the following formula:

Urine osmolality and specific gravity were measured in a morning specimen after eight-hour water deprivation. Urine-specific gravity was measured using an Assistant Urinprober hydrometer (model 242, Sondheim/Rhon, Germany), and values above 1005 were considered normal. The urine osmolality, determined cryoscopically, was considered to be normal when the following age-related values were reached: 850 mOsm/kg for persons ≤20 years old, 800 mOsm/kg for those aged 21–40 years, 700 mOsm/kg for those aged 40–60 years, and 600 mOsm/kg for persons ≥60 years old.[10] Anemia was defined as hemoglobin below 130 g/L for males and post-menopausal women and below 120 g/L for pre-menopausal women, as proposed by the World Health Organization.[11]

Kidney sonography was performed by one dedicated doctor using a Vivid 3-General Electric ultrasound machine with sector probe of 3.5 MHz. Cranio-caudal measurements of length and cross-sectional measurements of width and depth were made on each kidney in the sections visually estimated to represent the largest diameter. The measurements were expressed in centimeters. Kidneys of length below 10 cm were considered shrunken. Kidney volume was calculated using the ellipsoid formula:

A kidney volume below 124 cm³ and kidney volume BSA below 61 cm³/m² were considered to be reduced (i.e., below the values calculated as [arithmetic mean – SD]) for each of these parameters measured in healthy persons.

Statistical Analysis

Descriptive statistics are reported as frequency for categorical data and mean and standard deviation for continuous data. The comparison of variables among the four groups was made with one-way analysis of variance (ANOVA). The statistical significance of the differences between the groups was evaluated using a χ² test and unpaired t-test as appropriate. A value of p < 0.05 was considered significant. Sensitivity, specificity, and predictive values were calculated as described by Vecchio.[12]

All demographic, clinical, laboratory, and sonography variables determined in the study (i.e., both those included in the existing diagnostic criteria and the additional ones) were combined as independent variables in a univariate/multivariate logistic regression analysis. Epidemiological variables were not included in the logistic regression. Age was used as a continuous variable. For other variables, the presence of a pathologic value for each variable was recorded as 1 and the presence of a normal value as 2. Pathologic values were defined previously. Patients with diabetes mellitus were excluded from the analysis of the predictive value of glucosuria. Odds ratios for the presence of BEN, suspected BEN, and other kidney diseases, or the absence of any disease, were calculated.

RESULTS

The mean values of demographic, clinical, laboratory, and sonography parameters analyzed in the study are presented and compared in Table 1. There was no significant difference in sex, BSA, and BMI between the four groups examined. Healthy subjects and BEN-suspected patients were significantly younger than the patients of the two other groups. All three groups of patients had similar mean urine protein levels, urine-specific gravity, frequency of glucosuria, and kidney volume, but they were significantly different from the values for healthy persons. Also, the patient groups had significantly higher mean serum urea, creatinine, and urine alpha 1-microglobulin levels, lower creatinine clearance and urine osmolality, and shorter kidney length than healthy persons. These values were most greatly altered in BEN patients and significantly different in comparison with BEN-suspected and patients with other kidney diseases. BEN patients had significantly higher systolic blood pressure and lower mean hemoglobin level in comparison with all other groups. No difference in FENa, TRP, and urine alkaline phosphatase was found between the examined groups.

Table 1 Demographic, clinical, laboratory and sonography variables for the four examined groups

Variable	1. BEN	2. BEN suspected	4. Other	3. Healthy control	Statistical significance
Sex, F/M	10/9	11/12	17/17	9/14	ns
Age, year	68.3 ± 13.9	59.2 ± 13.5	67.9 ± 9.5	49.6 ± 12.8	p < 0.001
BSA, m^2	1.8 ± 0.2	1.9 ± 0.3	1.8 ± 0.2	1.9 ± 0.2	ns
BMI, kg/m^2	29.9 ± 6.0	28.1 ± 4.0	28.0 ± 4.1	27.9 ± 5.3	ns
Systolic BP, mmHg	176 ± 23	159 ± 25	155 ± 25	150 ± 28	p < 0.01
Diastolic BP, mmHg	100 ± 9	97 ± 15	92 ± 12	95 ± 17	p = ns
Danilovic's criteria
u-protein, g/24h	354 ± 259	331 ± 158	315 ± 230	112 ± 23	p < 0.01
Specific gravity	1015 ± 5	1017 ± 6	1016 ± 5	1019 ± 5	p < 0.02
Hemoglobin, g/L	129 ± 21	137 ± 15	132 ± 18	140 ± 8.0	p < 0.008
s- Urea, mmol/L	10.0 ± 4.2	8.0 ± 2.8	8.2 ± 5.3	6.0 ± 1.0	p < 0.01
s- Creatinine, umol/L	125 ± 45	88 ± 37	92 ± 52	77 ± 9	p < 0.01
Kidney length, cm	9.6 ± 0.6	10.3 ± 0.6	10.7 ± 0.9	11.5 ± 0.8	p < 0.0001
Other indices
Ccr mL/min	64.8 ± 29.5	90.0 ± 24.6	88.9 ± 24.9	115.6 ± 25.4*	p < 0.001
u-glucose, pts no	8	11	16	2	p < 0.025
u-alpha 1-MG, mg/mmolCr	6.94 ± 6.75	2.85 ± 4.77	2.17 ± 2.07	0.27 ± 0.35	p < 0.001
FENa,%	1.03 ± 0.45	1.16 ± 1.50	0.95 ± 0.47	0.69 ± 0.27	ns
TRP, %	82 ± 7	81 ± 31	84 ± 14	88 ± 4	ns
u-Osm, mOsm/L	661 ± 161	749 ± 185	700 ± 207	822 ± 147	p < 0.025
u-AP, U/mmol Cr	0.22 ± 0.12	0.32 ± 0.17	0.42 ± 0.22	0.18 ± 0.14	ns
Kidney volume, cm^3	100 ± 16	109 ± 29	102 ± 26	164 ± 42	p < 0.0001
Kidney volume/BSA, cm^3/m^2	56 ± 8	62 ± 14	55 ± 16	79 ± 18	p = 0.001

Abbreviations: BP = blood pressure, Ccr = creatinine clearance, alpha 1-MG = alpha 1-microglobulin, FENa = fractional sodium excretion, TRP = tubular phosphate reabsorption, u-OSm = urine osmolality, uAP = urine alkaline phosphatase.

*p < 0.05 as compared to all other groups.

The prevalence of each of the seven diagnostic criteria defined by Danilovic in the four examined groups is presented in Figure 1. All examined persons were from the BEN-affected village of Vreoci, and all BEN and BEN-suspected patients as well as 12/23 healthy controls belonged to BEN families. None of the remaining diagnostic criteria was found in healthy controls. The prevalence of proteinuria, anemia, and low specific gravity differed insignificantly among patients with BEN, BEN-suspected, and other kidney diseases. Azotemia and reduced kidney length were found in the majority of BEN patients but significantly less frequently in BEN-suspected patients and patients with other kidney diseases.

Figure 1. Prevalence of the diagnostic criteria defined by Danilovic in the four examined groups. Criteria: (1) farmers in the endemic villages, (2) familial history positive for BEN, (3) proteinuria, (4) low urine specific gravity, (5) anemia, (6) azotemia, (7) symmetrically shrunken kidneys.[1] *p < 0.05 as compared to all other groups (χ² –test).

Calculating the sensitivity and specificity of Danilovic's criteria to detect patients with BEN and suspected BEN revealed that the first criterion (residence in an endemic village) was nonspecific, while a positive family history had high sensitivity and specificity only in differentiating between BEN and BEN-suspected patients from those with other kidney diseases (see Table 2). Proteinuria had high sensitivity in differentiating between BEN and BEN-suspected patients from healthy persons and patients with other kidney diseases, but its specificity was low in differentiating between BEN and BEN-suspected patients from patients with other kidney diseases. Azotemia had high sensitivity in differentiating between BEN and other groups, but its specificity was high only in differentiating BEN from healthy persons. Low urine-specific gravity and shrunken kidneys exhibited high specificity but low sensitivity in differential diagnosis between BEN/BEN-suspected patients and healthy persons as well as patients with other kidney diseases (see Table 2).

Table 2 Sensitivity (%) and specificity (%) of the diagnostic criteria for Balkan endemic nephropathy proposed by Danilovic

	BEN vs. healthy		BEN vs. other		sBEN vs. healthy		sBEN vs. other		BEN vs. sBEN
Criteria	Sensitivity	Specificity	Sensitivity	Specificity	Sensitivity	Specificity	Sensitivity	Specificity	Sensitivity	Specificity
Farmers in BEN village	100	0	100	0	100	0	100	0	100	0
Familial history positive	100	47.8	100	100	100	47.8	100	100	100	0
Proteinuria	89.5	100	89.5	34.4	96.0	100	96.0	34.3	89.5	4.0
Low urine-specific gravity	33.0	100	33.0	85.3	8.00	100	8.00	85.3	27.8	79.2
Anemia	57.9	100	57.9	52.9	32.0	100	32.0	52.4	57.9	95.5
Azotemia	94.7	100	94.7	61.8	40.0	100	40.0	61.8	94.7	60.0
Shrunken kidney	68.4	100	68.4	84.2	16.0	100	16.0	74.2	68.4	84.0

In addition to Danilovic's criteria, demographic (sex, age), clinical (blood pressure), laboratory (creatinine clearance, glycosuria, alpha 1 microglobulinuria, urine alkaline phosphatase, %FENa, TRP, u-osmolality) and sonography (kidney length, kidney volume, kidney volume/BSA) variables were analyzed in a univariate logistic regression model in order to find those that significantly differentiate patients with BEN and suspected BEN from healthy persons and patients with other kidney diseases. The analyses showed that glucosuria and urine osmolality were not significant discriminatory variables, while low specific gravity appeared as a significant predictor of BEN in combination with healthy controls and BEN-suspected patients. All of the remaining variables appeared as significant predictive variables in the analysis of BEN/BEN-suspected patients and healthy persons as well as patients with other kidney diseases.

The variables found to have significant predictive value in univariate logistic regression analysis were used in the multivariate analysis. This showed that proteinuria was the only significant independent predictor for differentiation of BEN and BEN-suspected patients from healthy persons. Azotemia was a significant predictor of BEN in combination with BEN-suspected patients and azotemia and alpha 1 microglobulinuria in the combination with patients with other kidney diseases (see Table 3).

Table 3 Results of multivariate logistic regression analysis

Variable	B	S.E.	p	OR	95% CI
BEN vs. healthy
Proteinuria	5.41	1.46	.0002	224.00	12.79–3923.56
Constant	−8.05	2.31	.0005
sBEN vs. healthy
Proteinuria	5.92	1.45	.0000	373.99	21.78–6421.44
Constant	−9.0153	2.2892	.0001
BEN vs. other
Azotemia	3.33	1.14	.0034	27.85	3.01–257.72
Alpha1-MG	1.58	.80	.0459	4.86	1.03–22.93
Constant	−6.19	1.87	.0009
sBEN vs. other
Age	.17	.08	.0375	1.18	1.01–1.39
Proteinuria	6.95	3.25	.0325	1042.54	1.79–608607.83
Kidney volume	−3.13	1.48	.0348	.04	.002–.80
Constant	−13.52	8.11	.0955
BEN vs. sBEN
Azotemia	2.82	1.12	.0118	16.80	1.87–150.91
Constant	−3.16	1.33	.0176

Abbreviations: B = logistic regression coefficient, SE = standard error of regression coefficient, OR = odds ratio, CI = confidence interval, sBEN = BEN suspected.

DISCUSSION

Diagnosis of BEN is one of the problems always present in a discussion on this disease. Danilović, who was the first to describe BEN in Serbia, proposed some criteria for BEN diagnosis,[1] which have been used in many subsequent studies.[5],[8],[13] Authors from Zagreb used proteinuria, anemia, increased serum creatinine level, and a positive family history as diagnostic criteria, and they classified persons from endemic foci into three groups: diseased subjects, suspected subjects, and at risk subjects.[4],[14] Stefanović proposed that a diagnosis of BEN be made in patients with tubular abnormalities, proteinuria of the tubular type, scarce urinary deposit, absence of persistent urinary infection, and symmetric reduction of the kidneys with apparently normal pelvic and calyceal systems.[2],[6]

Although the specificity and sensitivity of these diagnostic criteria have not been examined, many authors have emphasized their inadequacy.[7],[15] In order to provide the most accurate diagnosis, some authors have recently used a combination of the criteria proposed in earlier studies.[16], [–18]

We considered that an evaluation of existing diagnostic criteria might contribute to studies on this problem. Therefore, in this work, we have calculated the prevalence, specificity, and sensitivity of Danilovic's criteria together with their predictive value in combination with several additional indices. The investigation was carried out on a sample taken from a large scale study in the BEN-affected village of Vreoci. Moreover, all the participants (BEN and BEN-suspected patients, those with other kidney diseases, as well as the healthy individuals) were all examined at the same time in the same institution.

An analysis of the prevalence of diagnostic criteria proposed by Danilovic showed that almost all BEN patients had azotemia, and its prevalence was significantly higher in BEN than in patients with suspected BEN or other kidney diseases. That explained the significantly more frequent reduced kidney length in BEN than in the two other patient groups (see Figure 1). Thus, this analysis as well as a comparison of the mean values for other parameters (see Table 1) has confirmed the fact that the current criteria enable diagnoses of BEN only in late stages of the disease. These patients had chronic kidney failure and therefore exhibited anemia, significantly higher values for urea and creatinine, lower creatinine clearance and kidney length in comparison with all the other groups (see Table 1). An examination of the sensitivity and specificity of the diagnostic criteria similarly showed that azotemia gave high sensitivity and specificity in differentiating between BEN patients and healthy individuals but not between BEN-suspected patients and healthy persons. Namely, chronic renal failure was less frequent in our BEN-suspected patient group, and therefore anemia, azotemia, and shrunken kidneys had sensitivities between 58 and 95% and specificities between 60 and 96% in differential diagnosis between BEN and BEN- suspected patients. Multivariate logistic regression also showed that azotemia was the only predictor for BEN vs. BEN-suspected patients, confirming that BEN-suspected patients differed from BEN only regarding kidney function. The present study also showed that patients designated as BEN-suspected according to the definition of Danilovic had numerous disorders of kidney function. In addition to proteinuria, many of them exhibited different disorders of proximal tubular function (glucosuria, alpha 1 microglobulinuria, decreased TRP, increased FENa), and a few of them had impaired urine-concentrating capacity, azotemia, and reduced kidney size. All of these indicated that these patients were not just BEN-suspected but BEN-diseased, and among them, some even had chronic renal failure. Therefore, definition of this group requires revision as well.

Calculations of the sensitivity and specificity of Danilovic's diagnostic criteria showed that only proteinuria had high sensitivity and specificity in distinguishing between both BEN and suspected BEN patients and healthy persons. However, it could be expected that neither proteinuria nor other criteria provided high sensitivity and specificity for a differential diagnosis between both BEN and BEN-suspected and other kidney diseases. Low specific gravity, anemia, azotemia, and shrunken kidneys are criteria of kidney failure and therefore have insufficient sensitivity and specificity in differential diagnosis between BEN and other kidney diseases, especially between suspected BEN and other kidney diseases. The multivariate analysis confirmed these results and showed that alpha 1 microglobulinuria was the only significant independent predictor for differentiation of BEN and patients with other kidney diseases. Although proteinuria and azotemia were showed as the variables with predictive values, they were also involved in Danilovic's diagnostic criteria for BEN and BEN-suspected, and therefore the values of odds ratio for these variables were high and 95% confidence intervals wide. However, exclusion of these variables from the model failed to discover any new independent predictor.

The Danilovic included symmetrically shrunken kidneys among his diagnostic criteria. However, before the appearance of ultrasonography as a non-invasive method for imaging, kidney size was rarely measured, and therefore other authors did not involve kidney size in diagnostic criteria.[3],[4] Recently, Dimitrov et al.[19] published the results of ultrasound measurements of the kidney in adult offspring of BEN families with normal creatinine clearance. They found shorter kidney length and smaller cortex in BEN offspring than in offspring of non-BEN families. This is in accordance with our previous study indicating that the process of shrinkage may begin in the early stadium of the disease.[13] Several authors indicated kidney volume, both absolute and relative, to be a more accurate measure of kidney mass that correlates better with glomerular filtration rate than kidney length.[20],[21] Therefore, kidney volume calculated by the ellipsoid formula and kidney volume/BSA were included in the present study beside kidney length, the most frequently used measure of kidney size. However, logistic regression failed to confirm that kidney volume was more powerful than kidney length in differentiating BEN and BEN-suspected patients from healthy persons and those with other kidney diseases.

A limitation of the present study is the relatively small patient number that prevents a completely reliable conclusion. Nevertheless, the results presented show that Danilovic's criteria had insignificant diagnostic value. They enabled the detection of BEN in advanced phase of the disease and only differential diagnosis between both BEN-patients and BEN-suspected patients and healthy persons. However, Danilovic's criteria had insufficient sensitivity and specificity in differential diagnosis between both BEN and suspected BEN and other kidney diseases. The indices of proximal tubular disorders involved in the present study were not found to be significant predictors of BEN, except for alpha 1 microglobulinuria, which discriminated significantly between BEN and other kidney diseases.

ACKNOWLEDGMENT

This work was supported by the Ministry of Science and Ecology of Serbia, contract 145037.