Acta Diabetol (1998) 35: 199–202
© Springer-Verlag 1998
ORIGINAL
I. Fernández Fernández · J. M. Páez Pinto T. Hermosín Bono · P Vázquez Garijo M. A. Ortiz Camuñez · M. A. Tarilonte Delgado
Rapid screening test evaluation for microalbuminuria in diabetes mellitus
Received: 12 September 1997 / Accepted in revised form: 4 September 1998
Abstract Microalbuminuria predicts clinical nephropathy and cardiovascular disease in diabetes mellitus. This study was undertaken to evaluate a screening microalbuminuria test with the Micral test II dipstick in a general practice setting and compare whether, if three urine samples are tested, any advantage is offered over a testing single sample. Two hundred and eighty diabetic patients attending a primary health care centre were studied. The first morning urine albumin concentration was determined by dipstick over 3 consecutive days. We studied two valuation methods: Method 1. Three-sample method: the test was considered positive if albumin was equal to or above 20 mg/l in at least two of three tests; Method 2. Singlesample method: we selected the third test, i.e. the most recent urine sample; if albumin was equal to or above 20 mg/l it was considered positive. The gold standard was the albumin excretion rate measured by a nephelometric method in a 24-h urine collection. Sensitivity, specificity, predictive values and Kappa coefficient were calculated. The diagnostic performance was assessed by a receiver operating characteristic curve. Microalbuminuria was defined for different thresholds of albumin excretion rate, 15, 20, 25 and 30 µg/min: their frequency was 38, 29, 23 and 18%, respectively. For method 1, the sensitivity of Micral test II oscillated between 70–94% and the specificity between 93–83%. For method 2, the sensitivity oscillated between 64–86%, and the specificity between 88–80%. Both methods had a high diagnostic performance. The Kappa coefficient was 84 and 60% for method 1 and 2, respectively. The Micral test II is a rapid, valid and reliable method for microalbuminuria screening in diabetic patients. It constitutes an important tool for diabetic surveillance in general practice. Although the use of three samples provides better results, the use of a single sample produces acceptable results at a low cost. I. Fernández Fernández (½) · J. M. Páez Pinto T. Hermosín Bono · P. Vázquez Garijo M. A. Ortiz Camuñez · M. A. Tarilonte Delgado Centro de Salud de Camas, C/Santa María de Gracia, 54, Camas, E-41900 Seville, Spain
Key words Microalbuminuria · Dipstick · Diabetes mellitus · General practice
Introduction
Microalbuminuria predicts persistent proteinuria, renal failure and cardiovascular disease in diabetes mellitus [1–3]. All diabetic patients should be screened yearly for early detection [4]. Microalbuminuria has been defined as an elevated urinary albumin value above the reference range for healthy subjects, but clinically undetectable by usual tests for proteinuria. An albumin excretion rate (AER) between 20–200 µg/min or 30–300 mg/24 h has been advocated as an operational definition. Radioimmunoanalysis (RIA), nephelometry and immunoturbidimetry are laboratory techniques used to determine microalbuminuria and have been shown to comply with the requirements of precision and analytic ranges [5]. However, they are not readily available, particularly for general practitioners (GPs). Furthermore, it is an advantage to have tests available which provide an approximation to the patient, i.e. obtain immediate results from the patient’s routine clinical care. The aims of our study were: 1) to evaluate the validity and reliability of the Micral test II, a new dipstick for the rapid detection of microalbuminuria in diabetes mellitus in a general practice setting, and 2) to compare whether, by using three urine samples, any advantage is offered over a single sample.
Subjects, materials and methods An observational study on diagnostic tests was carried out. Our hypothesis was that the Micral test II dipstick is a valid test for microalbuminuria screening, both with three and with single samples. We selected 333 patients (132 were men, 201 were women), consecutively attending at a primary care health centre, including 18 with type 1 (insulin-dependent) and 315 with type 2 (non-insulin-
200
J. Fernández Fernández et al.: Test for microalbuminuria in diabetes mellitus
Fig. 1 Subjects
dependent) diabetes. These patients were attending for routine diabetes care. Their mean ±SD ages and duration of diabetes were 57 ±6 and 10 ±4 years, respectively. All the patients were examined for factors which may affect the test (heart failure, acute febrile illnesses, urinary infections). Of the 333 subjects 53 were excluded, 29 because they showed leukocytes, red blood cells or nitrites on Multistix testing (Ames) and 24 because they had an AER 200 µg/min or more and were thus regarded as having “clinical proteinuria”. Therefore, only 280 diabetic patients were included for microalbuminuria study (Fig. 1). Patients were instructed to collect a sample of the first urine passed on three consecutive mornings. The samples were stored at +4 °C, until day 3 when each sample was tested by a general practitioner with Micral test II (Boehringer-Mannheim) dipsticks according to the manufacturer’s instructions, i.e. submerging the dipstick in the urine for 5 s, and taking the reading 1 min later. The Micral test II was compared with a reference colour scale which discriminates the negative value and provides a semiquantitative (0, 20, 50 and 100 mg/l) measure. When the Micral test II read 100 mg/l, 1 ml urine was diluted with 1 ml water and the reading of the urinary albumin concentration was doubled. We compared two methods: 1) three-sample method: if two of the three consecutive urine Micral test II readings were equal to or above 20 mg/l, the test was considered positive for microalbuminuria. 2) Single-sample method: we selected the most recent urine sample, i.e. the third test, which was considered positive if the reading was equal to or above 20 mg/l. As the gold standard, a nephelometric measure of the albumin excretion rate (AER) was carried out on three 24-h urine samples collected during the following 4–6 weeks from all patients. Microalbuminuria was defined for different thresholds of AER, 15, 20, 25 and 30 µg/min. Persistent microalbuminuria was considered if two of the three measures of AER were equal to or above the defined threshold. The measures of the gold standard and the methods studied were blinded. Validity was studied by the sensitivity, specificity and positive and negative predictive values. Interobserver variability was assessed by the Kappa coefficient (%), with two blinded observers. The diagnostic performance of Micral test II was illustrated by means of drawing ROC (Receiver Operating Characteristic) curves. The abscissa of the curve is the fraction of patients with AER less than 20 µg/min who are false positives with Micral test II. The ordinate is the fraction of patients with AER 20 µg/min or more who have been correctly classified. The perfect classification rule has an ROC curve that follows the left and top borders on the graph.
Results
We studied the clinical proteinuria separately. All 24 (100%) patients had a positive test for detection of clinical proteinuria using the three-sample method, whereas 22 (92%) had a positive test using the single-sample method. Of the 280 remaining patients, 81 (29%) had microalbuminuria (i.e., they had an AER between 20 and 200 µg/ min). Employing other thresholds of the AER in order to ascertain the presence of microalbuminuria, it was found that 108 patients (38%) had an AER between 15 and 200 µg/min, 64 (23%) had an AER between 25 and 200 µg/min, and 51 (18%) had an AER between 30 and 200 µg/min. Table 1 shows the data for sensitivity, specificity, and predictive values for the utilisation of three urine samples considering different AER thresholds. Likewise, the area below the ROC curve is presented as a diagnostic performance rate of the Micral test II. Table 2 demonstrates the results for the utilisation of just one urine sample. Figure 2 shows the ROC curves for three-sample and single-sample methods. Both methods have a high diagnostic performance (area under the ROC curves). The Kappa coefficient for interobserver variability was 84% in the three-sample analysis and 60% in the single-sample analysis.
Discussion
Our results show that the Micral test II dipstick is valid and reliable for the detection of albuminuria (microalbuminuria and clinical proteinuria) in diabetic patients, the test resulting in a high diagnostic performance. The period samples (24-h old or from night urine) continue to provide the reference model, but they do tend to be uncomfortable for the patient – which could cause an error factor in clinical practice which could affect the interpretation of the results. Using a simple method is of great use in the identification of patients at risk.
J. Fernández Fernández et al.: Test for microalbuminuria in diabetes mellitus Table 1 Micral test II evaluation using three urine samples and different thresholds for definition of microalbuminuria
201
Threshold (µg/min)
Sensitivity (%) Specificity (%) Predictive value + (%) Predictive value – (%) Correct classifications (%) Area under ROC curve (95% CI)
15–200 n = 108
20–200 n = 81
25–200 n = 64
30–200 n = 51
70 93 86 83 84 0.85 (0.80–0.90)
84 90 77 93 88 0.90 (0.86–0.95)
91 87 67 97 88 0.91 (0.86–0.96)
94 83 55 99 85 0.93 (0.89–0.98)
CI, Confidence interval; ROC, Receiver Operating Characteristic Table 2 Micral test II evaluation using a single urine sample and different thresholds for definition of microalbuminuria
Threshold (µg/min)
Sensitivity (%) Specificity (%) Predictive value + (%) Predictive value – (%) Correct classifications (%) Area under ROC curve (95% CI)
15–200 n = 108
20–200 n = 81
25–200 n = 64
30–200 n = 51
64 88 78 80 79 0.82 (0.77–0.88)
79 87 72 91 85 0.89 (0.84–0.94)
83 83 60 94 83 0.88 (0.89–0.94)
86 80 49 96 81 0.90 (0.85–0.95)
CI, Confidence interval; ROC, Receiver Operating Characteristic Fig. 2 Receiver operating characteristic (ROC) curves: screening test for detection of urinary albumin excretion ≥ 20 µ/min. The abscissa of the curve is the fraction of patients with AER < 20 µg/min who were false positives with Micral test II. The ordinate is the fraction of patients with AER ≥ 20 µg/min who were correctly classified. The perfect classification rule has an ROC curve that follows the left and top borders on the graph
The albumin/creatine ratio has been utilised for random urine samples or for the first morning urine sample as a valid and simple trial for selective exploration in population studies and in routine clinical practice. Some authors question whether more information is obtained from the determination of the albumin/creatine ratio in comparison to the sole determination of albumin concentration [6, 7]. Various authors have compared qualitative techniques based on the specific union of albumin to a bromophenol colouring (Micro-Bumintest) with laboratory techniques, obtaining good correlations and satisfactory reliability [8–10] for AER above 40 mg/l. Subsequently, the Micral test dipstick appears to be the most simple, efficient and fastest method for the semiquantitative detection of microalbuminuria with excellent correlations with laboratory techniques, and high sensitivity and specificity for discriminating concentrations of urine albumin of 20–30 mg/l
[11–15]. Recently, Micral test II has been commercialised, which shortens the time necessary for the technique, thus facilitating the determination even more. The determination time has been reduced to 1 min – giving a read-off margin of between 1 and 5 min. Mogensen et al. [16], considering thresholds of 15 and 20 mg/l in order to establish the abnormal albumin concentration, observed with Micral test II, respectively, a sensitivity of 95 and 97%, a specificity of 97 and 72%, a positive predictive value of 97 and 84% and a predictive negative value of 94 and 88%, the proportion of our correctly classified samples being 95 and 87%. Gilbert et al. [17], in comparison with a radioimmunoassay show a sensitivity of 93%, specificity of 93% and positive predictive value of 89%. Kutter et al. [18] refer to a correlation coefficient of 0.938. Our results, although relatively acceptable, do not appear to be so optimal, probably due to the utilisation of a different gold standard: we
202
J. Fernández Fernández et al.: Test for microalbuminuria in diabetes mellitus
used the AER (µg/min) whereas they used AC (mg/l) to define the microalbuminuria. Poulsen et al. [19] evaluated three different users in the utilisation of the dipsticks: 1) qualified nurses, 2) laboratory technicians, and 3) general practitioners. They reached the conclusion that with qualified nurses and laboratory technicians sensitivity improves compared to the general practitioners (91, 84, and 66%). They explain that the low sensitivity of the general practitioner is probably due to their lack of experience and incorrect handling of the dipsticks. In our study a general practitioner performed all the determinations, improving on the sensitivity found by Poulsen, which shows the importance of training. For the detection of “clinical proteinuria”, the use of specific dipsticks for microalbuminuria also improves upon the results of the conventional dipsticks for albuminuria (36–73% sensitivity) [20, 21] and the Micro-Bumintest reactive tablets (87%) [20]. Our results with the Micral test II dipsticks were 100% sensitivity with three samples and 92% with one sample. In our experience the Micral test II dipsticks are a rapid, valid and reliable method for microalbuminuria screening in diabetic patients. The best results are obtained for an AER threshold value of 20 µg/min and using three dipsticks, as recommended by the manufacturer, but the use of just one dipstick demonstrates sufficiently valid results and is easier and more economical. The screening cost was $ 4.5 and $ 1.5 per patient in the three samples and single sample method, respectively. We consider, as did Le Floch et al. [22], that the dispstick is cost-effective for annual microalbuminuria screening in diabetic patients. The presence of microalbuminuria alerts the physician to the need for improved diabetic control and other risk factors. Our findings suggest that Micral test II is a useful tool for microalbuminuria screening and constitutes an important tool for diabetic surveillance in general practice. However, as Gilbert et al. [17] claimed, it should not be regarded as a diagnostic test: a positive result should be followed-up by measuring AER. Although our results are orientated to their usefulness in the diabetic patient, we think that their use could be extended and applied to hypertensive patients, as a vascular damage indicator, and to pregnant women as a possible predictor in the development of preeclampsia [13]. Acknowledgements We wish to thank Daniel Pasca Rey for designing the computer programme to perform ROC curve calculations. Dr. José Mialdea, former resident family and community medicine physician at the Camas Health Centre, for her collaboration in the reliability evaluation test. Rosario Morato, Celia Rivero, Francisco Javier Iglesias, Modesto Caro and the rest of the medical, nursing, auxiliary and administrative staff of the Camas Health Centre, for their collaboration and help which has enabled us to finish this task. The Fondo de Investigación Sanitaria (Health Research Fund) for funding the project (94/0952).
References 1. Viberti GC, Hill RD, Jarret RJ, Argyropoulos A, Mahmud U, Keen H (1982) Microalbuminuria as a predictor of clinical nephropathy in insulin-dependent diabetes mellitus. Lancet I: 1430–1432
2. Mattock MB, Morrish NJ, Viberti GC, Keen H, Fitzgerald AP, Jackson G (1992) Prospective study of microalbuminuria as a predictor of mortality in NIDDM. Diabetes 41: 736–741 3. Alzaid AA (1996) Microalbuminuria in patients with NIDDM: an overview. Diabetes Care 19: 79–89 4. Krans HMJ, Porta M, Keen H (1992) Diabetes care and research in Europe: The Saint Vincent Declaration Action Programme. World Health Organization. Copenhagen, pp 29–32 5. Poulsen PL (1994) Microalbuminuria: técnicas de determinación. In: Mogensen CE (ed) Microalbuminuria: un marcador de lesión orgánica. Science Press London, pp 11–20 6. Howwey JEA, Browning MCK, Fraser CG (1989) Biologic variations of urinary albumin: consequences for analysis, specimen collection, interpretation of results and screening programs. Am J Kidney Dis 13: 35–37 7. Zelmanovitz T, Gross JL, Oliveira JR, Paggi A, Tatsch M, Azevedo MJ (1997) The receiver operating characteristics curve in the evaluation of a random urine specimen as a screening test for diabetic nephropathy. Diabetes Care 20: 516–521 8. Slama G, Boillot J, Desplanque N, Letanoux M (1985) Estimación de mciroalbuminuria a la cabecera del enfermo. Lancet (Ed. Esp) 7: 94 9. Tai J, Tze WJ (1990) Evaluation of Micro-Bumintest reagent tablets for screening of microalbuminuria. Diabetes Res Clin Pract 9: 137–142 10. Vernetta I, Porto MA, Berengue i Iglesias MD, Alvarez i Funes V, Juvanteny i Gorgals J, Cespedes Lamarca A (1990) Detección de microalbuminuria mediante tabletas reactivas en la DMNID: validación de la técnica en atención primaria. Atención Primaria 7: 482–486 11. Bangstad HJ, Try K, Dahl-Jorgensen K, Hanssen KF (1991) New semiquantitative dipstick test for microalbuminuria. Diabetes Care 14: 1094–1097 12. Jury DR, Mikkelsen DJ, Glen D, Dunn PJ (1992) Assessment of Micral-test microalbuminuria test strip in the laboratory and in diabetic outpatients. Ann Clin Biochem 29: 96–100 13. Hasslacher C (1993) Clinical significance of microalbuminuria and evaluation of the Micral-test. Clin Biochem 26: 283–287 14. Tiu SC, Lee SS, Cheng MW (1993) Comparison of six commercial techniques in the measurement of microalbuminuria in diabetic patients. Diabetes Care 16: 616–620 15. De Grauw WJ, van de Lisdonk EH, van de Hoogen WH, van Gerwen WH, van den Bosch WJ, Willems JL, van Weel C (1995) Screening for microalbuminuria in type 2 diabetic patients: the evaluation of a dipstick test in general practice. Diabet Med 12:657–663 16. Mogensen CE, Vestbo E, Poulsen PL, Christiansen C, Damsgaard EM, Eijskjaer A, Hansen KW, Nielsen S, Pedersen MM (1995) Microalbuminuria and potential confounders. A review and some observations on variability of urinary albumin excretion. Diabetes Care 18: 572–581 17. Gilbert RE, Akdeniz A, Jerums G (1997) Detection of microalbuminuria in diabetic patients by urinary dipstick. Diabetes Res Clin Pract 35: 57–60 18. Kutter D, Thoma J, Hansen S, Carl R (1995) Screening for oligoalbuminuria by means of Micral-Test II. A new immunological test strip. Eur J Clin Chem Clin Biochem 33: 243–245 19. Poulsen PL, Hansen B, Amby T, Terkelsen T, Mogensen CE (1992) Evaluation of a dipstick test for microalbuminuria in three different clinical settings, including the correlation with urinary albumin excretion rate. Diabete Metab 18: 395–400 20. Higby K, Suiter CR, Siler-Khodr T (1995) A comparison between two screening methods for detection of microproteinuria. Am J Obstet Gynecol 173: 1111–1114 21. Olivarius ND, Mogensen CE (1995) Danish general practitioners’ estimation of urinary albumin concentration in the detection of proteinuria and microalbuminuria. Br J Gen Pract 45: 71–73 22. Le Floch JP, Charles MA, Philipson C, Perlemuter L (1993) Costeffectiveness of screening for microalbuminuria using immunochemical dipstick tests or laboratory assay in diabetic patients. Diabet Med 11: 349–356