- Email: [email protected]
Inaccuracies in digoxin measurement
Clin Btochem, Vol. 21, pp. 353-357, 1988
0009-9120/88 $3.00 + .00 Copyright © 1988 The Canadian Society of Clinical Chemists.
Printed in Canada. All rights reserved.
Inaccuracies in Digoxin Measurement JANINE
DENIS
C O O K , 1 T H O M A S R. K O C H , 1 M I C H A E L and E D W A R D C. K N O B L O C K 3
S. C O O K , 2
1 D e p a r t m e n t of P a t h o l o g y , U n i v e r s i t y of M a r y l a n d S c h o o l of M e d i c i n e , 10 S. P i n e Street, B a l t i m o r e , M D 2 1 2 0 1 , and 2 R e s e a r c h and T r a i n i n g P r o f e s s i o n a l s , B a l t i m o r e , M D 2 1 2 1 8 Six commercial digoxin immunoassaykits were evaluatedfor their accuracy of calibration and their extent of interference by digoxinlikeimmunoreactivesubstance (DLIS). Calibration accuracy was investigatedwith digoxin referencestandards in pooled human serum. The Abbott and Becton Dickinson kits underestimatewhile the other kits overestimatedigoxin concentration. The magnitude of this bias generally increases with increasing concentration of digoxin. Sera from digoxin-free patient populations with potential DLIS interference-pregnant women, newborns, hypertensives, and uremicsn wereanalyzedwith each kit. Healthy subjects not on digoxin therapy sewed as controls. Groups with DLIS interference, as exemplified by a significant difference of p < 0.05 from controls, are: Abbott-newborns and pregnant women; Becton Dickinson--newborns and pregnant women; Dade--no difference; Dupont--newborns, uremics,pregnant women, and hypertensives;Kallestad--newborns; and Syva--newborns. The limitations of each individual digoxin method should be realizedfor DLIS interferenceand bias, and patient results from that method should be interpreted accordingly.
ies considerably among immunoassay kits, most likely due to different binding affinities of DLIS with the commercial digoxin antibodies (3). We report two studies designed to investigate bias of digoxin tests. To study accuracy of calibration, digoxin reference standards in pooled human serum were prepared and analyzed. To investigate the extent of DLIS interference with each kit, we tested serum from digoxin-free patient populations with potential DLIS interference. Serum samples from pregnant women, newborns, hypertensives, and uremics were analyzed. Additionally, sera from controls and patients receiving digoxin therapy were tested.
KEY WORDS: digoxin; digoxin-like immunoreactive substance (DLIS); cross-reactivity; immunoassay.
MATERIALS
Introduction
D
igoxin is a glycosylated steroid-like drug derived from foxglove which is indicated in the treatment of congestive heart failure, atrial fibrillation, or atrial flutter. Its mode of action involves the reversible inhibition of the Na+,K+-ATPase pump leading to initiation of cardiac contractions. Because of digoxin's narrow therapeutic index, assays should be reproducible and free of significant bias. Careful monitoring of serum drug levels is important to avoid toxicity manifested as potentially life-threatening arrhythmias. Most common digoxin assay methods are immunoassays, which afford the sensitivity necessary for the detection of nanogram levels of digoxin (1). Despite the excellent sensitivity, these methods are not absolutely specific. The most important cross-reactive substance in human serum to interfere with immunological dig0xin analysis is termed digoxin-like immunoreactive substance (DLIS) (2). The cross reactivity of DLIS vatCorrespondence: Thomas R. Koch, Ph.D., The University of MarylandHospital, Clinical Laboratories, 22 S. Greene Street, Baltimore, MD 21201. 3Deceased October 7, 1987. Manuscript received May 2, 1988; accepted May 9, 1988. CLINICALBIOCHEMISTRY, VOLUME 21, DECEMBER 1988
Materials a n d m e t h o d s
Absolute methanol: Burdick and Jackson Laboratories, Inc., Muskegon, MI 49442. USP reference standard digoxin: United States Pharmacopiel Convention, Inc., Rockville, MD 20852. REFERENCE STANDARDS We prepared 1.3, 2.6, and 3.2 nmol/L digoxin reference standards by adding 0.100, 0.200, and 0.250 mL ofdigoxin stock solution (2.56 umol/L in absolute methanol) to three separate 200 mL portions of serum pool. We pooled filtered serum from blood donors found to be negative for hepatitis B surface antigen and HIV antibody. These pools were stored at -20°C in 1.5 mL capped polypropylene tubes (Brinkman Instruments Co., Westbury, NY 11590). SUBJECTS For controls (n = 10) we chose healthy ambulatory subjects aged 23-65. The digoxin group (n = 10) consisted of hospitalized patients currently receiving digoxin. To test for DLIS interference, we selected digoxinfree subjects from several groups: uremia (n = 8)-hospitalized patients with serum creatinine > 619 umol/L; hypertension (n = 12)--outpatients attending the hypertension clinic; pregnancy (n = 1 7 ) outpatients > 36 weeks of gestation; newborns (n = 353
DENIS COOK, KOCH, COOK, AND KNOBLOCK
11)--blood was aspirated by syringe from the fetal cord vein. For each subject in the study, we pooled leftover serum and stored 1.5 mL portions in polypropylene tubes at - 20°C.
ugal analyzer (Cobas Bio: Roche Diagnostics Systems, Inc., Nutley, NJ 07110).
METHODS
Data were analyzed using the SPSS/PC + statistics package (4).
All assays were performed according to the manufacturer's instructions. Controls recommended by each manufacturer were used to evaluate the success of each run. For each manufacturer, a single lot of reagents was used. Abbott TDx Digoxin II: Abbott Diagnostics, Inc., North Chicago, IL 60064. Becton Dickinson Digoxin Solid Phase Component System: Becton Dickinson and Co., Orangebury, NY 10962. Dade Stratus Digoxin Fluorometric Enzyme Immunoassay: American Dade, Miami, FL 33152. Dupont Digoxin: E. I. du Pont de Nemours and Co., Wilmington, DE 19898. Kallestad Quantitope Digoxin Radioimmunoassay: Kallestad Laboratories, Inc., Austin, TX 78701. Syva EMIT Column Digoxin Assay: Syva Co., Palo Alto, CA 94303. We performed this assay on a centrif-
i5
.8
-G
J~
5
"o
¢
o
113
~
~® =
STATISTICS
Results REFERENCE STANDARDS
We analyzed the digoxin reference standards as unknowns in multiple runs with each method. Those methods which used stored calibration (Abbott, Dade, Dupont, and Syva) showed no calibration drift during the course of these experiments. From these data we assessed the accuracy of calibration and precision of each method (Figure 1). The Dupont kit was the most precise method, as exemplified by between-run % C.V., while the Kallestad kit was the least precise. Based on these data, we would predict that the Abbott and Becton Dickinison methods underestimate and the others overestimate digoxin concentrations. For most
~
o
R J~
130"
T
120-
~
I10-
~
90-
>
80-
0 U
°
-I-
--I.p
--b
z-'- 1 +
"-II
I'
I
I
' 7O
,I
I
I
C X
!
~
o
I |
a I
I (21} (3Sl (21)i(12) (3{ (20)
1,5 nrnol//
(22) (3s) (2z)[(~3) (~s] (201
2.6 nmol/L
(2(
[381 (20) 03) (~4B F'o)
3.2 nmol/L
Digoxin Concentrotion F i g u r e 1 - - R e s u l t s for Reference S t a n d a r d s . Recovery of digoxin for reference s t a n d a r d s . Precisions, as exemplified by betweenr u n % C.V., are s h o w n as e r r o r bars. V a l u e s in p a r e n t h e s e s r e p r e s e n t t h e n u m b e r of d e t e r m i n a t i o n s . L i n e a r regression analyses of each k i t w i t h t h e t h r e e levels of spiked h u m a n s e r u m pool gave t h e following results: Abbott, y = - 0.03 + 0.88x (r = 0.98); Becton Dickinson, y = - 0 . 0 5 + 0.94x (r = 0.98); Dade, y = - 0 . 0 6 + 1.10x (r = 0.98); Dupont, y = 0.40 + 0.93x (r = 0.97); K a l l e s t a d , y = 0.01 + 1.13x (r = 0.85); a n d Syva, y = - 0 . 1 5 + 1.10x (r = 0.96). 354
CLINICAL BIOCHEMISTRY, VOLUME 21, DECEMBER 1988
INACCURACIES IN DIGOXIN MEASUREMENT methods, bias increases with increasing concentration of digoxin. Two-way analysis of variance revealed a significant difference in bias between the different kits (F = 45.7, p < 0.0001). Furthermore, kits differ in their accuracy of measuring digoxin at the three specified concentrations (F = 3.6, p < 0.001). Irrespective of digoxin concentration, the mean overall error per test is: Abbott, - 0 . 3 1 nmol/L (n = 63); Becton Dickinson, - 0.19 nmol/L (n = 114); Dade, 0.15 nmol/L (n = 63); Dupont, 0.24 nmol/L (n = 38); Kallestad, 0.32 nmol/L (n = 114); and Syva, 0.09 nmol/L (n = 60). This bias for each kit across all three digoxin pool levels was significantly different by analysis of variance (F = 42.6, p < 0.0001). We applied the Duncan multiple range procedure to assess which kits differ significantly from one another (p < 0.05) across all digoxin concentrations. All kits differ from one another, with the following exceptions: Dade did not differ significantly from Dupont and Syva; Kallestad did not differ from Dupont; Abbott and Becton Dickinson did not differ from each other. DLIS INTERFERENCE Analysis of specimens from various groups of subjects (Figure 2) showed major differences for the different methods. By analysis of variance and the Duncan multiple range procedure we compared the patient groups
with normal controls, as measured by each individual kit. With the Abbott TDx II method, we found high results for newborns. In addition, results in pregnant women were higher t h a n in uremics and normal controls. Newborns differed significantly from all other groups while pregnant women differed from controls, hypertensives, and uremics with the Becton Dickinson method. The smallest difference among groups was observed with the Dade method. We found falsely elevated digoxin measurements for all groups with the Dupont method. In spite of this constant bias, newborns were found to differ from all other groups while uremics, pregnant women, and hypertensives were significantly different from controls with the Dupont method. With the Kallestad method newborns differed significantly from all other groups. Relatively minor differences were observed with the Syva method, although, newborns measured significantly higher t h a n other groups. Patients with liver failure are also subject to DLIS interference with digoxin assays. Though not a part of our original protocol, we analyzed eight patients with serum total bilirubin ranging from 39.3 to 268.5 umol/L by the Abbott and Dade digoxin methods. The two methods did not differ significantly from each other, but both gave results t h a t were significantly different (p < 0.05) from their respective normal controls. We also tested newborn (n = 3) and uremic (n = 3)
4.5 ¸
3.8-
~1
.c x o ,~
2.6
1.9
1.3 ¸
o6
O.O
ttlI
~
-
Abbott
Becton Dickinson
Dode
Dupont
Kotlestod
Syva
Figure 2--Results for subjects on digoxin therapy (digoxin) or not receiving digoxin (all other groups). The dots represent the mean, the error bars the range, and the cross hatch marks the 95% confidence interval of the mean for each group with each kit. CLINICAL BIOCHEMISTRY, VOLUME 21, DECEMBER 1988
355
DENIS COOK, KOCH, COOK, AND KNOBLOCK patients who were receiving digoxin. For both groups, there was a significant difference between serum digoxin values obtained with different kits (newborns: F = 23.0, p < 0.01: uremics: F = 150.7, p < 0.001).
Discussion Digoxin is one of the most widely prescribed drugs in North America. Because of its narrow therapeutic index (1.2-2.6 nmol/L), and because the clinical signs and symptoms of either digoxin overdose or noncompliance are not specific, dosing must be based on serum drug concentration measurements. Results for reference standards indicate that digoxin values from different methods may agree poorly. Some kits underestimate and others overestimate digoxin concentration, and the magnitude of the bias increases with increasing concentration. The bias found between different digoxin methods is unacceptable in light of the drug's narrow therapeutic range. This variability could jeopardize physician management of digitalized patients when different assays are used to monitor therapy. For instance, at the clinically significant decision level of 2.6 nmol/L, the Abbott and Kallestad methods would differ on average by 0.8 nmol/L. To eliminate DLIS interference, a digoxin immunoassay must either have an antibody so specific for digoxin that DLIS will not cross-react or include a pretreatment step to separate DLIS from digoxin prior to assay (5). The degree of interference caused by DLIS varies considerably with the immunoassay kit used. This variability most likely results from the different binding affinities of DLIS with the different digoxin antibodies. The kit antibodies appear to have higher specificity for the steroid portion of the digoxin molecule rather than the carbohydrate end, which is altered during labeling (1). Greater specificity for the carbohydrate portion could decrease DLIS cross-reactivity. Also, individual assay protocols may affect the degree of DLIS interference encountered. In normal persons > 90% of serum DLIS is tightly but reversibly bound to serum proteins. As such it is not easily detectable with most digoxin immunoassays. In uremics, hypertensives, pregnant women, and newborns, Valdes (6) has suggested that DLIS becomes weakly bound or unbound from serum proteins and thus more readily detectable. These patients are thought to have an expanded extracellular volume and thus stimulated DLIS production (7). Volume expansion and also increasing levels of extracellular sodium supposedly stimulate the production of endoxin, the natriuretic hormone considered to be DLIS, which inhibits Na ÷, K+-ATPase. In our study, we found no difference between uremic patients and controls when tested with the modified Abbott TDX digoxin assay. In contrast, Branchi (8) found that the serum of uremic patients undergoing long-term dialysis treatment had a high background fluorescence, most likely due to endogenous fluorescent "uremic toxins" which randomly rotate within the fluorescent field. 356
With the three patients evaluated in our study that were both uremic and on digoxin therapy, a high background blank was encountered which necessitated further dilution of their sera. Results for uremic patients differed (p < 0.05) from controls only with the Dupont method. Pregnant women with cardiac disease are often digitalized to aid in their clinical management. Because pregnancies complicated with cardiac problems have significant maternal mortality (9), maintenance of therapeutic digoxin levels is imperative. Also, the mother can be digitalized to treat the fetus with perinatal heart failure (10). Serum DLIS increases throughout gestation (6), but then rapidly decreases postpartum to reach undetectable levels by 24 h after delivery. This suggests that the postpartum serum half-life for DLIS is six h or less (11). Four of the methods tested exhibited significant difference (p < 0.05) between controls and pregnant women. DLIS causes the greatest interference in the newborn population. In premature infants DLIS peaks four to six days after birth and usually disappears by two months of age (12). Some authors have recommended measuring a baseline digoxin concentration on neonates prior to use of digoxin, and subtraction of the baseline result from subsequent results (13,14). This approach is inadequate because the bias is not linearly related to age. Our results with fetal cord blood serum agree with those of Gault et al. (15) and Ng et al. (16) who found significant cross reactivity in serum from low birth weight infants with radioimmunoassay and fluorescence polarization immunoassay methods, but minimal interference with radial fluorescence immunoassay. The degree of interference with the Abbott TDx renders that method useless with neonatal samples. Hypertensives are thought to have endoxin production due to expanded extracellular volume. Only the Dupont method exhibited a significant difference (p < 0.05) between hypertensives and controls. For adequate management of patients receiving digoxin, the limitations of each individual method in use should be well understood in terms of DLIS interference in certain population groups and the inherent bias of the method. A reference method for digoxin measurement should be established and a protocol devised whereby manufacturers can achieve a degree of standardization in the calibration of methods (17). Until methods come into common usage which are accurately calibrated and free from DLIS interference, it is essential for laboratories to advise clinicians of the limitations of existing methods.
Acknowledgements This study was sponsored by the U.S. Food and Drug Administration under an interagency agreement with the Baltimore Veterans Administration Medical Center Research Service. The authors wish to thank the manufacturers whose kits were evaluated in this study (Abbott, Becton Dickinson, Dade, Dupont, Kallestad, and Syva) for their generous donation of all necessary reagents. CLINICAL BIOCHEMISTRY, VOLUME 21, DECEMBER 1988
INACCURACIES IN DIGOXIN MEASUREMENT
Dedication This manuscript is dedicated to the late Colonel Edward C. Knoblock. It was his unfailing commitment to the training of future clinical chemists that made this research project possible.
References 1. Soldin SJ, Papanastasiou-Diamandi A, Heyes J, Lingwood C, Olley P. Are immunoassays for digoxin reliable? Clin Biochem 1984; 17: 317-20. 2. Besch Jr HR, Hufferd S, Lake M, Hurwitz R, Watanabe AM. False elevation of apparent digoxin levels in plasma of premature infants. Clin Chem 1976; 22: 1168. 3. Pudek MR, Seccombe DW, Jacobson BE, Whitfield MF. Seven different digoxin immunoassay kits compared with respect to interference by digoxin-like immunoreactive substance in serum from premature and full-term infants. Clin Chem 1983; 29: 1972-4. 4. Norusis MJ. S P S S / P C +. Chicago: SPSS, Inc., 1985. 5. Skogen WF, Rea MR, Valdes Jr R. Endogenous digoxinlike immunoreactive factors eliminated from serum samples by hydrophobic silica-gel extraction and enzyme iramunoassay. Clin Chem 1987; 33: 401-4. 6. Valdes Jr R. Endogenous digoxin-like immunoreactive factor: impact on digoxin measurements and potential physiological implications. Clin Chem 1985; 31: 1525-32. 7. Grantham JJ, Edwards RM. Natriuretic hormones: at last, bottled in bond? J Lab Clin Med 1984; 103: 333-6.
CLINICAL BIOCHEMISTRY, VOLUME 21, DECEMBER 1988
8. Bianchi P. Interferences in TDx digoxin assay in dialysis patients. Clin Chem 1986; 32: 2099. 9. Ureland K. What's the risk when the cardiac patient is pregnant? Contemp Obstet Gynecol 1979; 13: 117-20. 10. Kerenyi TD, Gleicher N, Miller J, Brown E, Steinfeld L, Chitkara U, Raucher H. Transplacental cardioversion of intrauterine supraventricular tachycardia with digitalis. Lancet 1980; 2: 393-5. 11. Graves SW, Valdes Jr R, Brown BA, Knight AB, Craig HR. Endogenous digoxin-immunoreactive substance in human pregnancies. J Clin Endocrinol Metab 1984; 58: 748-51. 12. Valdes Jr R, Graves SW, Brown BA, Landt M. Endogenous substances in newborn infant causing false positive digoxin measurements. J. Pedriatr 1983; 102: 947-50. 13. McCarthy RC. Minimizing the effect of digoxin-like immunoreactive substances in immunoassays for digoxin in neonatal serum. Clin Chem 1985; 31: 1240-1. 14. Pudek MR, Seccombe DW, Whitfield MF, Digoxin-like immunoreactivity in premature and full-term infants not receiving digoxin therapy. N Engl J Med 1983; 308: 9045. 15. Gault MH, Vasdev S, Longerich L. Higher values for digitalis-like factors with TDx Digoxin II. Clin Chem 1986; 32: 2000-1. 16. Ng PK, LeGatt D, Coates J, Collins-Nakai RL. Measuring endogenous digoxin-like substance and exogenous digoxin in serum of low-birth weight infants. A m J Hosp Pharm 1985; 42: 1977-9. 17. Clark DR, Inloes RL, Kalman SM, Sussman HH. Abbott TDx, Dade Stratus, and Du Pont aca automated digoxin immunoassays compared with a reference radioimmunoassay method. Clin Chem 1986; 32: 381-5.
357
0009-9120/88 $3.00 + .00 Copyright © 1988 The Canadian Society of Clinical Chemists.
Printed in Canada. All rights reserved.
Inaccuracies in Digoxin Measurement JANINE
DENIS
C O O K , 1 T H O M A S R. K O C H , 1 M I C H A E L and E D W A R D C. K N O B L O C K 3
S. C O O K , 2
1 D e p a r t m e n t of P a t h o l o g y , U n i v e r s i t y of M a r y l a n d S c h o o l of M e d i c i n e , 10 S. P i n e Street, B a l t i m o r e , M D 2 1 2 0 1 , and 2 R e s e a r c h and T r a i n i n g P r o f e s s i o n a l s , B a l t i m o r e , M D 2 1 2 1 8 Six commercial digoxin immunoassaykits were evaluatedfor their accuracy of calibration and their extent of interference by digoxinlikeimmunoreactivesubstance (DLIS). Calibration accuracy was investigatedwith digoxin referencestandards in pooled human serum. The Abbott and Becton Dickinson kits underestimatewhile the other kits overestimatedigoxin concentration. The magnitude of this bias generally increases with increasing concentration of digoxin. Sera from digoxin-free patient populations with potential DLIS interference-pregnant women, newborns, hypertensives, and uremicsn wereanalyzedwith each kit. Healthy subjects not on digoxin therapy sewed as controls. Groups with DLIS interference, as exemplified by a significant difference of p < 0.05 from controls, are: Abbott-newborns and pregnant women; Becton Dickinson--newborns and pregnant women; Dade--no difference; Dupont--newborns, uremics,pregnant women, and hypertensives;Kallestad--newborns; and Syva--newborns. The limitations of each individual digoxin method should be realizedfor DLIS interferenceand bias, and patient results from that method should be interpreted accordingly.
ies considerably among immunoassay kits, most likely due to different binding affinities of DLIS with the commercial digoxin antibodies (3). We report two studies designed to investigate bias of digoxin tests. To study accuracy of calibration, digoxin reference standards in pooled human serum were prepared and analyzed. To investigate the extent of DLIS interference with each kit, we tested serum from digoxin-free patient populations with potential DLIS interference. Serum samples from pregnant women, newborns, hypertensives, and uremics were analyzed. Additionally, sera from controls and patients receiving digoxin therapy were tested.
KEY WORDS: digoxin; digoxin-like immunoreactive substance (DLIS); cross-reactivity; immunoassay.
MATERIALS
Introduction
D
igoxin is a glycosylated steroid-like drug derived from foxglove which is indicated in the treatment of congestive heart failure, atrial fibrillation, or atrial flutter. Its mode of action involves the reversible inhibition of the Na+,K+-ATPase pump leading to initiation of cardiac contractions. Because of digoxin's narrow therapeutic index, assays should be reproducible and free of significant bias. Careful monitoring of serum drug levels is important to avoid toxicity manifested as potentially life-threatening arrhythmias. Most common digoxin assay methods are immunoassays, which afford the sensitivity necessary for the detection of nanogram levels of digoxin (1). Despite the excellent sensitivity, these methods are not absolutely specific. The most important cross-reactive substance in human serum to interfere with immunological dig0xin analysis is termed digoxin-like immunoreactive substance (DLIS) (2). The cross reactivity of DLIS vatCorrespondence: Thomas R. Koch, Ph.D., The University of MarylandHospital, Clinical Laboratories, 22 S. Greene Street, Baltimore, MD 21201. 3Deceased October 7, 1987. Manuscript received May 2, 1988; accepted May 9, 1988. CLINICALBIOCHEMISTRY, VOLUME 21, DECEMBER 1988
Materials a n d m e t h o d s
Absolute methanol: Burdick and Jackson Laboratories, Inc., Muskegon, MI 49442. USP reference standard digoxin: United States Pharmacopiel Convention, Inc., Rockville, MD 20852. REFERENCE STANDARDS We prepared 1.3, 2.6, and 3.2 nmol/L digoxin reference standards by adding 0.100, 0.200, and 0.250 mL ofdigoxin stock solution (2.56 umol/L in absolute methanol) to three separate 200 mL portions of serum pool. We pooled filtered serum from blood donors found to be negative for hepatitis B surface antigen and HIV antibody. These pools were stored at -20°C in 1.5 mL capped polypropylene tubes (Brinkman Instruments Co., Westbury, NY 11590). SUBJECTS For controls (n = 10) we chose healthy ambulatory subjects aged 23-65. The digoxin group (n = 10) consisted of hospitalized patients currently receiving digoxin. To test for DLIS interference, we selected digoxinfree subjects from several groups: uremia (n = 8)-hospitalized patients with serum creatinine > 619 umol/L; hypertension (n = 12)--outpatients attending the hypertension clinic; pregnancy (n = 1 7 ) outpatients > 36 weeks of gestation; newborns (n = 353
DENIS COOK, KOCH, COOK, AND KNOBLOCK
11)--blood was aspirated by syringe from the fetal cord vein. For each subject in the study, we pooled leftover serum and stored 1.5 mL portions in polypropylene tubes at - 20°C.
ugal analyzer (Cobas Bio: Roche Diagnostics Systems, Inc., Nutley, NJ 07110).
METHODS
Data were analyzed using the SPSS/PC + statistics package (4).
All assays were performed according to the manufacturer's instructions. Controls recommended by each manufacturer were used to evaluate the success of each run. For each manufacturer, a single lot of reagents was used. Abbott TDx Digoxin II: Abbott Diagnostics, Inc., North Chicago, IL 60064. Becton Dickinson Digoxin Solid Phase Component System: Becton Dickinson and Co., Orangebury, NY 10962. Dade Stratus Digoxin Fluorometric Enzyme Immunoassay: American Dade, Miami, FL 33152. Dupont Digoxin: E. I. du Pont de Nemours and Co., Wilmington, DE 19898. Kallestad Quantitope Digoxin Radioimmunoassay: Kallestad Laboratories, Inc., Austin, TX 78701. Syva EMIT Column Digoxin Assay: Syva Co., Palo Alto, CA 94303. We performed this assay on a centrif-
i5
.8
-G
J~
5
"o
¢
o
113
~
~® =
STATISTICS
Results REFERENCE STANDARDS
We analyzed the digoxin reference standards as unknowns in multiple runs with each method. Those methods which used stored calibration (Abbott, Dade, Dupont, and Syva) showed no calibration drift during the course of these experiments. From these data we assessed the accuracy of calibration and precision of each method (Figure 1). The Dupont kit was the most precise method, as exemplified by between-run % C.V., while the Kallestad kit was the least precise. Based on these data, we would predict that the Abbott and Becton Dickinison methods underestimate and the others overestimate digoxin concentrations. For most
~
o
R J~
130"
T
120-
~
I10-
~
90-
>
80-
0 U
°
-I-
--I.p
--b
z-'- 1 +
"-II
I'
I
I
' 7O
,I
I
I
C X
!
~
o
I |
a I
I (21} (3Sl (21)i(12) (3{ (20)
1,5 nrnol//
(22) (3s) (2z)[(~3) (~s] (201
2.6 nmol/L
(2(
[381 (20) 03) (~4B F'o)
3.2 nmol/L
Digoxin Concentrotion F i g u r e 1 - - R e s u l t s for Reference S t a n d a r d s . Recovery of digoxin for reference s t a n d a r d s . Precisions, as exemplified by betweenr u n % C.V., are s h o w n as e r r o r bars. V a l u e s in p a r e n t h e s e s r e p r e s e n t t h e n u m b e r of d e t e r m i n a t i o n s . L i n e a r regression analyses of each k i t w i t h t h e t h r e e levels of spiked h u m a n s e r u m pool gave t h e following results: Abbott, y = - 0.03 + 0.88x (r = 0.98); Becton Dickinson, y = - 0 . 0 5 + 0.94x (r = 0.98); Dade, y = - 0 . 0 6 + 1.10x (r = 0.98); Dupont, y = 0.40 + 0.93x (r = 0.97); K a l l e s t a d , y = 0.01 + 1.13x (r = 0.85); a n d Syva, y = - 0 . 1 5 + 1.10x (r = 0.96). 354
CLINICAL BIOCHEMISTRY, VOLUME 21, DECEMBER 1988
INACCURACIES IN DIGOXIN MEASUREMENT methods, bias increases with increasing concentration of digoxin. Two-way analysis of variance revealed a significant difference in bias between the different kits (F = 45.7, p < 0.0001). Furthermore, kits differ in their accuracy of measuring digoxin at the three specified concentrations (F = 3.6, p < 0.001). Irrespective of digoxin concentration, the mean overall error per test is: Abbott, - 0 . 3 1 nmol/L (n = 63); Becton Dickinson, - 0.19 nmol/L (n = 114); Dade, 0.15 nmol/L (n = 63); Dupont, 0.24 nmol/L (n = 38); Kallestad, 0.32 nmol/L (n = 114); and Syva, 0.09 nmol/L (n = 60). This bias for each kit across all three digoxin pool levels was significantly different by analysis of variance (F = 42.6, p < 0.0001). We applied the Duncan multiple range procedure to assess which kits differ significantly from one another (p < 0.05) across all digoxin concentrations. All kits differ from one another, with the following exceptions: Dade did not differ significantly from Dupont and Syva; Kallestad did not differ from Dupont; Abbott and Becton Dickinson did not differ from each other. DLIS INTERFERENCE Analysis of specimens from various groups of subjects (Figure 2) showed major differences for the different methods. By analysis of variance and the Duncan multiple range procedure we compared the patient groups
with normal controls, as measured by each individual kit. With the Abbott TDx II method, we found high results for newborns. In addition, results in pregnant women were higher t h a n in uremics and normal controls. Newborns differed significantly from all other groups while pregnant women differed from controls, hypertensives, and uremics with the Becton Dickinson method. The smallest difference among groups was observed with the Dade method. We found falsely elevated digoxin measurements for all groups with the Dupont method. In spite of this constant bias, newborns were found to differ from all other groups while uremics, pregnant women, and hypertensives were significantly different from controls with the Dupont method. With the Kallestad method newborns differed significantly from all other groups. Relatively minor differences were observed with the Syva method, although, newborns measured significantly higher t h a n other groups. Patients with liver failure are also subject to DLIS interference with digoxin assays. Though not a part of our original protocol, we analyzed eight patients with serum total bilirubin ranging from 39.3 to 268.5 umol/L by the Abbott and Dade digoxin methods. The two methods did not differ significantly from each other, but both gave results t h a t were significantly different (p < 0.05) from their respective normal controls. We also tested newborn (n = 3) and uremic (n = 3)
4.5 ¸
3.8-
~1
.c x o ,~
2.6
1.9
1.3 ¸
o6
O.O
ttlI
~
-
Abbott
Becton Dickinson
Dode
Dupont
Kotlestod
Syva
Figure 2--Results for subjects on digoxin therapy (digoxin) or not receiving digoxin (all other groups). The dots represent the mean, the error bars the range, and the cross hatch marks the 95% confidence interval of the mean for each group with each kit. CLINICAL BIOCHEMISTRY, VOLUME 21, DECEMBER 1988
355
DENIS COOK, KOCH, COOK, AND KNOBLOCK patients who were receiving digoxin. For both groups, there was a significant difference between serum digoxin values obtained with different kits (newborns: F = 23.0, p < 0.01: uremics: F = 150.7, p < 0.001).
Discussion Digoxin is one of the most widely prescribed drugs in North America. Because of its narrow therapeutic index (1.2-2.6 nmol/L), and because the clinical signs and symptoms of either digoxin overdose or noncompliance are not specific, dosing must be based on serum drug concentration measurements. Results for reference standards indicate that digoxin values from different methods may agree poorly. Some kits underestimate and others overestimate digoxin concentration, and the magnitude of the bias increases with increasing concentration. The bias found between different digoxin methods is unacceptable in light of the drug's narrow therapeutic range. This variability could jeopardize physician management of digitalized patients when different assays are used to monitor therapy. For instance, at the clinically significant decision level of 2.6 nmol/L, the Abbott and Kallestad methods would differ on average by 0.8 nmol/L. To eliminate DLIS interference, a digoxin immunoassay must either have an antibody so specific for digoxin that DLIS will not cross-react or include a pretreatment step to separate DLIS from digoxin prior to assay (5). The degree of interference caused by DLIS varies considerably with the immunoassay kit used. This variability most likely results from the different binding affinities of DLIS with the different digoxin antibodies. The kit antibodies appear to have higher specificity for the steroid portion of the digoxin molecule rather than the carbohydrate end, which is altered during labeling (1). Greater specificity for the carbohydrate portion could decrease DLIS cross-reactivity. Also, individual assay protocols may affect the degree of DLIS interference encountered. In normal persons > 90% of serum DLIS is tightly but reversibly bound to serum proteins. As such it is not easily detectable with most digoxin immunoassays. In uremics, hypertensives, pregnant women, and newborns, Valdes (6) has suggested that DLIS becomes weakly bound or unbound from serum proteins and thus more readily detectable. These patients are thought to have an expanded extracellular volume and thus stimulated DLIS production (7). Volume expansion and also increasing levels of extracellular sodium supposedly stimulate the production of endoxin, the natriuretic hormone considered to be DLIS, which inhibits Na ÷, K+-ATPase. In our study, we found no difference between uremic patients and controls when tested with the modified Abbott TDX digoxin assay. In contrast, Branchi (8) found that the serum of uremic patients undergoing long-term dialysis treatment had a high background fluorescence, most likely due to endogenous fluorescent "uremic toxins" which randomly rotate within the fluorescent field. 356
With the three patients evaluated in our study that were both uremic and on digoxin therapy, a high background blank was encountered which necessitated further dilution of their sera. Results for uremic patients differed (p < 0.05) from controls only with the Dupont method. Pregnant women with cardiac disease are often digitalized to aid in their clinical management. Because pregnancies complicated with cardiac problems have significant maternal mortality (9), maintenance of therapeutic digoxin levels is imperative. Also, the mother can be digitalized to treat the fetus with perinatal heart failure (10). Serum DLIS increases throughout gestation (6), but then rapidly decreases postpartum to reach undetectable levels by 24 h after delivery. This suggests that the postpartum serum half-life for DLIS is six h or less (11). Four of the methods tested exhibited significant difference (p < 0.05) between controls and pregnant women. DLIS causes the greatest interference in the newborn population. In premature infants DLIS peaks four to six days after birth and usually disappears by two months of age (12). Some authors have recommended measuring a baseline digoxin concentration on neonates prior to use of digoxin, and subtraction of the baseline result from subsequent results (13,14). This approach is inadequate because the bias is not linearly related to age. Our results with fetal cord blood serum agree with those of Gault et al. (15) and Ng et al. (16) who found significant cross reactivity in serum from low birth weight infants with radioimmunoassay and fluorescence polarization immunoassay methods, but minimal interference with radial fluorescence immunoassay. The degree of interference with the Abbott TDx renders that method useless with neonatal samples. Hypertensives are thought to have endoxin production due to expanded extracellular volume. Only the Dupont method exhibited a significant difference (p < 0.05) between hypertensives and controls. For adequate management of patients receiving digoxin, the limitations of each individual method in use should be well understood in terms of DLIS interference in certain population groups and the inherent bias of the method. A reference method for digoxin measurement should be established and a protocol devised whereby manufacturers can achieve a degree of standardization in the calibration of methods (17). Until methods come into common usage which are accurately calibrated and free from DLIS interference, it is essential for laboratories to advise clinicians of the limitations of existing methods.
Acknowledgements This study was sponsored by the U.S. Food and Drug Administration under an interagency agreement with the Baltimore Veterans Administration Medical Center Research Service. The authors wish to thank the manufacturers whose kits were evaluated in this study (Abbott, Becton Dickinson, Dade, Dupont, Kallestad, and Syva) for their generous donation of all necessary reagents. CLINICAL BIOCHEMISTRY, VOLUME 21, DECEMBER 1988
INACCURACIES IN DIGOXIN MEASUREMENT
Dedication This manuscript is dedicated to the late Colonel Edward C. Knoblock. It was his unfailing commitment to the training of future clinical chemists that made this research project possible.
References 1. Soldin SJ, Papanastasiou-Diamandi A, Heyes J, Lingwood C, Olley P. Are immunoassays for digoxin reliable? Clin Biochem 1984; 17: 317-20. 2. Besch Jr HR, Hufferd S, Lake M, Hurwitz R, Watanabe AM. False elevation of apparent digoxin levels in plasma of premature infants. Clin Chem 1976; 22: 1168. 3. Pudek MR, Seccombe DW, Jacobson BE, Whitfield MF. Seven different digoxin immunoassay kits compared with respect to interference by digoxin-like immunoreactive substance in serum from premature and full-term infants. Clin Chem 1983; 29: 1972-4. 4. Norusis MJ. S P S S / P C +. Chicago: SPSS, Inc., 1985. 5. Skogen WF, Rea MR, Valdes Jr R. Endogenous digoxinlike immunoreactive factors eliminated from serum samples by hydrophobic silica-gel extraction and enzyme iramunoassay. Clin Chem 1987; 33: 401-4. 6. Valdes Jr R. Endogenous digoxin-like immunoreactive factor: impact on digoxin measurements and potential physiological implications. Clin Chem 1985; 31: 1525-32. 7. Grantham JJ, Edwards RM. Natriuretic hormones: at last, bottled in bond? J Lab Clin Med 1984; 103: 333-6.
CLINICAL BIOCHEMISTRY, VOLUME 21, DECEMBER 1988
8. Bianchi P. Interferences in TDx digoxin assay in dialysis patients. Clin Chem 1986; 32: 2099. 9. Ureland K. What's the risk when the cardiac patient is pregnant? Contemp Obstet Gynecol 1979; 13: 117-20. 10. Kerenyi TD, Gleicher N, Miller J, Brown E, Steinfeld L, Chitkara U, Raucher H. Transplacental cardioversion of intrauterine supraventricular tachycardia with digitalis. Lancet 1980; 2: 393-5. 11. Graves SW, Valdes Jr R, Brown BA, Knight AB, Craig HR. Endogenous digoxin-immunoreactive substance in human pregnancies. J Clin Endocrinol Metab 1984; 58: 748-51. 12. Valdes Jr R, Graves SW, Brown BA, Landt M. Endogenous substances in newborn infant causing false positive digoxin measurements. J. Pedriatr 1983; 102: 947-50. 13. McCarthy RC. Minimizing the effect of digoxin-like immunoreactive substances in immunoassays for digoxin in neonatal serum. Clin Chem 1985; 31: 1240-1. 14. Pudek MR, Seccombe DW, Whitfield MF, Digoxin-like immunoreactivity in premature and full-term infants not receiving digoxin therapy. N Engl J Med 1983; 308: 9045. 15. Gault MH, Vasdev S, Longerich L. Higher values for digitalis-like factors with TDx Digoxin II. Clin Chem 1986; 32: 2000-1. 16. Ng PK, LeGatt D, Coates J, Collins-Nakai RL. Measuring endogenous digoxin-like substance and exogenous digoxin in serum of low-birth weight infants. A m J Hosp Pharm 1985; 42: 1977-9. 17. Clark DR, Inloes RL, Kalman SM, Sussman HH. Abbott TDx, Dade Stratus, and Du Pont aca automated digoxin immunoassays compared with a reference radioimmunoassay method. Clin Chem 1986; 32: 381-5.
357