Problems of Interfering Substances Identified

Evaluation of Automated Urinary lodine Methods: Problems of Interfering Substances Identified

Warwick May,' Diana Wu,' Creswell Eastman,' Pierre Bourdoux,^ and Glen Maberly'

We evaluated automated methods for measurement of uri biochemical déterminations have advanced rapidly. nary iodine (Ul) over a range expected in iodinereplete and Our aim w£is to evaluate the performance of the auto iodinedeficient populations. Results obtained with Tectini mated Ul System available since 1973, which is based on con AutoAnalyzer II Systems, based on either dialysis or acid dialysis of urine (method A). Results from samples assayed digestion, were compared witfi those obtained by a manual at Westmead Hospital, Australia, were correlated with alkaline ashing techinique. Results of automated dialysis results fi^m the Saint Pierre University Hospital, Bel were consistently higher than those obtained by the other gium, which were obtained with an even earlier automated methods. The apparently higher concentrations of Ul we Ul détermination, based on acid digestion (method B). measured were due to interfering substances crossing the Results from thèse automated methods were also compared with results obtained by a more technically demanding dialysis membrane and participating in the catalytic reaction. manual technique (method G, performed at Westmead Thiocyanate (SCN) was one endogenous substance contrib Hospital), based on the oSicial method of the Association of uting to the increased measurement of Ul. For urinary SCN OflRcial Analytical Chemists (AOAC) (6), in which alkaline concentrations of 5 to 15 mg/L, the amount of overestimation ashing is used to destroy organic matter in the sample in the Ul measurement attributable to SCN ranged from 21.8 before the final colorimetric measurement of iodine. to 61 Atg/L. However, SCN may account for only 40-50% of the apparent increase in Ul. In samples with lower Ul (<50 Materials and Methods /ng/L), interfering substances produced a 100% error in Apparatus results. We conclude that the automated dialysis System should not be used to assess iodinedeficient populations. Automated measurement of Ul was performed with a This leaves a major dilemma for researchers wanting to Technicon AutoAnalyzer II system (Technicon Instrument assess the iodine status of populations, because the auto Corp., Tarrytown, NY) with either a dialysis module (meth od A) or acid digestion unit (method B). Ashing of urine mated digestion method is no longer commercially available. samples (method C) was carried out in a spécial 102C muffle fumace (Scientific Equipment Manufacturers, Addltlonal Keyphrases: dialysis and acid digestion techniques Magill, South Australia). To measure absorbance, we used compared • analytical error • thiocyanate • nutritional status a Spectronic 20 spectrophotometer (Milton Roy, Rochester, NY). Endémie goiter, resulting from nutritional iodine defi ciency, is one of the oldest recognized disorders aflfecting Reagents whole populations. Recently, Hetzel (1) proposed the term Glassdistilled deionized water, used to prépare ail solu "iodinedeficiency disorders" to reflect the broader spec tions and for dilutions, was passed through an ionex trum of this condition, the eflFects of which include sponta change colimm containing anionexchange resin AG501 neous abortion, increased infant mortality, and cretinism. X8(D) (BioRad, Richmond, CA) to remove any traces of In fact, the stunting of intellectual functioning in entire iodine. AU other reagents were of analytical grade and iodinedeficient communities has been demonstrated (2). were found to have negligible iodine content. The International Council for the Control of lodine Defi ciency Disorders has estimated that 800 million people Urine Samples Worldwide, living mostly in Asia, Africa, and South Amer Casual (untimed) urine samples from 51 patients and ica, are at risk of iodine-deficiency disorders (3). staff (Westmead Hospital) and from 11yearold schoolchil Accordingly, many countries have made or implemented dren (Tasmania, Australia) were selected to cover the plans to set up centers to assess and monitor iodine defi range of Ul concentrations found in the Australian popu ciency. This usually includes automated laboratories for lation. After collection, samples were acidified and refing measuring urinary iodine (Ul). Although Ul excrétion has erated until assayed. been considered the définitive criterion for grading the severity of iodine deficiency in communities (4), over the Procédures past two décades this test has become rare in clinical Alkaline ashing. Urine samples (3 mL) were treated with chemistry departments in the developed world. Indeed, 1 mL of 1 mol/L potassium hydroxide and ashed at 600 °C since modification of the earlier automated acid digestion according to the method of Belling (7). We measured the System in 1973 (5), there has been no real progress in iodine in the ashed solution by assessing the catalji;ic eflfect automated Ul meeisurement, whereas methods for other of iodine on the SandellKolthofi" reaction, as described by Potter et al. (8). Because the potassium hydroxide used for ashing the samples depresses the catalytic eflfect of iodine ' Australian Centre for the Control of lodine Disorders, West- on the SandellKolthoflf reaction, the calibration curve mead Hospital, Sydney, Australia. ^ Department of Nuclear Medicine, Saint Pierre University obtained with nonashed aqueous standards is always Hospital, Free University of Brussels, Brussels, Belginm. steejjer than the curve produced by adding iodine to ashed Received J^UEu-y 3, 1990; accepted March 21,1990. solutions. To compensate for this, we derived a correction

CLINICAL CHEMISTRY, Vol. 36, No. 6,1990 865 fadbr by adding known amounts of iodine (5 or 10 ng per doux et al. (13). tube), in triplicate, to a séries of urine and blank samples Data analysis. Statistical analysis—linear régression, (n = 17). In several assays (n = 9), we calculated the slope Pearson's corrélation coefficient (r)—was performed with of the line obtained by plotting the amoimt of added iodine StatView 512+™ (Brainpower Inc., Calabassa, CA) on an against the change in absorbance at 340 nm (a), and Apple Macintosh computer. compared this with the slope of a line similarly produced with use of nonashed aqueous standards (6). We deter- Results mined the correction factor (expressed as b:a) to be 1.31 ± Recovery and précision. The mean (± SD) analytical 0.08 (mean ± SD); i.e., values obtained for urine samples by recovery we obtained in method C, after the addition of interpolation from a standard curve prepared with use of radioisotopic and nonlabeled iodine, was 100.3 (2.9)% and aqueous standards must be multiplied by 1.31 to obtain the 97.8 (7.2)%, respectively. Thèse results are in close agree- correct resuit. ment with the analytical recoveries published by others Analytical recovery of iodine after alkaline ashing was who used this method (7,14). Similarly, method A gave a assessed in two ways. First, we prepared urine samples mean analytical recovery of 99% (SD 3.7%), which accords (n = 8) as usual, but added 15 IJL of [^^^I]NaI to give about favorably with the mean recovery of 97% obtained by Gary 70 000 coimts/min. We counted the radioactivity before and et al. (5). lodine-supplemented urine assayed by method B after the ashing procédure and expressed the final count as also showed an analytical recovery of 96-97% (authors' a percentage of the initial counts. We also assessed the unpublished data). analytical recovery of nonlabeled iodine from urine after The between-assay précision (CV) for the assay of a the ashing and assay stages as follows: 2,4, or 6 ng of iodine control urine with a mean UI concentration of 11.3 /^/L, in was added, in triplicate, to 3-niL aliquots of urine (n = 4). a total of 30 alkaline ashing assays, was 10.6%. In method After ashing, we determined the final concentration of A, the between-assay CV for the three urine controls with iodine in each tube as described above. Recovery was mean UI concentrations of 24, 45, and 94 /xg/L was 3.8%, calculated as the amount of added iodine measured, ex• 2.5%, and 2.4%, respectively (n = 35 assays). pressed as a percentage of the amount of iodine added. Discrepancy between methods. Figure 1 compares the UI Reproducibility was evtduated with use of a control urine results for the samples in each method. UI is plotted on a spécimen stored frozen in 3-mL aliquots. Aliquots of this log scale, with results of the alkaline ashing method used to control were then thawed, ashed, and assayed in duplicate rank samples in decreasing order. This Figure demon- with each run. strates an increase in results obtained in automated dial• Automated analysis of urinary iodine. The Technicon ysis, especially in samples with UI <200 jug/L. Régression continuous-flow system used for determining UI by method analysis further highlights the bias between the ashing B is a modification of the procédure used for measuring method (x) and automated dialysis (y), yielding a corréla• protein-bound iodine in sérum, in which an automated tion coefficient (r) of 0.94, but a régression équation oiy = digester hélix is used to digest the samples with concen- 0.91x + 62.9. On the other hand, the corrélation between trated minerai acids at 320 °C (9). UI measurements by ashing (x) and automated digestion (y) was very close for method A were performed with the automated system the range studied (r = 0.99, y = 0.93x + 7.9). described by Gary et al. (5), in which the digester unit, just Interfering substances. To try to explain the apparent described, is replaced by the simpler and safer dialysis overestimation in UI in method A, we investigated the module. Results were calculated by a personal computer possible eflFect of thiocyanate, an ion that might cross the programmed to correct for baseline drift and nonlinearity dialysis membrane and interfère with the Sandell-Kolthoff of the standard curve. reaction (9,15). Table 1 shows a significant increase in the Analytical recovery of iodine from three urine samples UI concentrations (after the addition of increasing amounts (chosen with initied UI concentrations équivalent to the of thiocyanate to urine) as measured in automated dialysis, low, medivmi, and high régions of the standard curve) was assessed by adding increasing concentrations of iodine « ALKALINE ASIlINQ standard (25-125 /xg/L) to an equal volume of urine. For quality control of the automated dialysis method, we assayed 1-mL aliquots of three urine S8unples with low, médium, and high iodine concentrations. Thèse quality- control samples were stored frozen, then thawed and as• sayed in duplicate with each set of 10-15 samples analyzed. Interférence of thiocyanate. To investigate the possible interférence of urinary thiocyanate (USCN) on UI détermi• nations by the différent methods, we added potassium thiocyanate solutions (prepared in iodine-free water) to various samples to give a final concentration range of 1.25-20 mg/L (21.5-344 /xmol/L), thus simulating the USCN concentrations conmionly found in nonsmokers, 4.9 ± 2.3 mg/L, as well as the higher concentrations seen in 10 H • ,—• 1—•• 1—• 1 , , , , smokers, 14.6 ± 7.1 mg/L (10). Concentrations of USCN 0 10 20 30 40 50 60 >15 mg/L are also common in populations consuming SAMPLE NUMBER cassava in their diet {10,11). We also determined the effect Fig. 1. Comparison of UI (/^g/L) values obtained by the three of endogenous USCN for each method. Where suflficient différent methods sample remained (n = 38), we determined USCN by using Alkaline ashing results used to rank samples in decreasing order. Note log a modified method of Aldridge (12) as described by Bour- scale of y-axis

866 CLINICAL CHEMISTRY, Vol. 36, No. 6,1990 y = 3.379X + 21.743 , r = 0.60 Table 1. Effect of Exogenous SCN on Iodine Déterminations SCN Mean (SD) Ul, fig/L SCN i by added,* added," digestion, mg/L Dialysis Aile, ashing mg/L 0" 16(0.3) 13.8 (2.9) 2.5 49 1.25 21 (0.4)" 16.4 (3.0) 5 51 2.5 26 (0.5)" 19(0.1) 7.5 50 5 38 (0.8)" 19.4 (2.7) 10 48 10 59(1.2)" 17.4 (0.8) 12.5 49 0 2 4 6 8 i: 12 14 16 18 20 22 24 "Added to an equal volume of urine. "Added to an equal volume of iodine standard. "Urine diluted with an equal volume H2O only (baseline). "Statis- UEC:f (rr.g/L) tically différent from baseline (P <0.0001). n = 3 each. Fig. 3. Corrélation between USCN concentration and the différence between Ul as measured by automated dialysis and by alkaline ashing but not in alkaline ashing. The two columns at the far right of Table 1 indicate an absence of interférence by SCN on iodine measurement in automated digestion, in which Table 2. Decrease in Ul Measured in Four Urine increasing concentrations of SCN solutions were mixed Samples by Automated Dialysis, after Ashing to with an equal volume of the 100 /xg/L iodine standard. Remove SCN Subséquent addition of increasing amounts of thiocyanate Before ashing After ashing to 15 urine samples produced a consistent response in the apparent UI concentration measured in method A (Figure SCN, mg/L Ui, /ig/L SCN, mg/L Ul, MQ/L 2). Régression analysis of thèse data gives the following 3.45 73 nd 40 équation: y = 3.93x + 2.2 (r = 0.99, P <0.0001). In other 5.00 61 nd 20 words, a mean USCN concentration of 5 mg/L (nonsmok- 4.3S @ nd 23 ers) or 15 mg/L (smokers or cassa va consumers) will cause 3.25 33 . nd 14.5 an overestimation in measured UI of 21.8 and 61 jag/L, respectively, when analyzed by automated dialysis. nd = not détectable, n = 3. The eflFect is also seen with endogenous USCN (Figure 3). y = 4,698x • 62.4 , r = 0 6/ The différence between UI measured in method A, minus 180, the UI concentration measured in method C, is signifi- cantly correlated (r = 0.6, P <0.0001) to the USCN concen• tration in each sample. This relationship was absent in method B, the régression équation generated for the différ• ence in UI (method B - method C) (y) and USCN (x) being 3- = 0.3x - 1.6 (r = 0.085, P = 0.62). Furthermore, as Table 2 shows, the UI concentration in four urine samples mea• sured by method A decreased after each sample was ashed to remove interfering substances such as SCN. 20 J , . , , , , , , , , . , Finally, this overestimation in the UI measurement 0 2 3 ;.0 l.Z 10 0 12.5 15 0 17 5 20.0 22 5 caused by USCN in method A is more critical in samples USC:J •rr;,l) with low UI, because as the true amoimt of iodine in a Fig. 4. Relationship between Ul concentrations determined by sample déclines, the proportion of apparent iodine mea• automated dialysis and USCN concentration for ail samples with Ul sured that is caused by thèse interfering substances in- <100 fjg/l (as determined by alkaline ashing) creases. Figure 4 illustrâtes a direct corrélation between the UI concentrations by method A and the USCN content significant for this relationship (P <0.0001). Furthermore, for samples with UI <100 jug/L (as determined by alkaline UI results obtained by method C and method B showed no ashing). The corrélation coefficient of 0.67 was highly significant corrélation with USCN content.

y = 3.925X » 2.;67 , r = 0.99 Discussion 100, Comparing the UI results in three methods, we identified a serious bias in the Technicon AutoAnalyzer II System based on sample dialysis (Figure 1). The automated method based on acid digestion and a manual alkaline ashing technique were in close agreement for the studied range. Because this différence could not be accounted for by any systematic iodine loss, we postulated that an interfering substance such as thiocyanate might account for the over• estimation in UI measured by the dialysis method. Addi• 3 2 4 6 3 :0 12 14 16 18 20 22 tion of thiocyanate solutions to samples assayed in each T:- ;CVAMATE (mg/D method showed a direct linear relationship between UI Fig. 2. Effect of added SCN on Ul results by the dialysis method measiu-ed by automated dialysis and the USCN content Each concentration of SCN was added to 15 urines (Figure 2), but not in the other two methods (Table 1).

CLINICAL CHEt^/IISTRY, Vol. 36, No. 6,1990 867 If endogenous thiocyanate behaves similarly, the différ• phasized: hazards include the potentially explosive nature ence between the UI concentration in automated dialysis of perchloric acid, and the libération of a considérable and alkaline ashing should be accounted for by the amount quantity of fumes, including toxic chlorine gas (18). Labo• of USCN in each sample. Results showed that such a ratories using acid digestion techniques should have adé• relationship did exist, but as seen in Figure 3, with the quate exhaust ventilation, preferably incorporating a fume ji-intercept value of 21.7 Mg/L, USCN alone does not ac- hood specifically designed for use with perchloric acid. count for the total différence between UI concentrations Similarly, accurate measurement of UI by the manual measured by the various methods. Thus other substemces alkaline ashing technique is difi&cult, labor intensive, and may also cross the dialysis membrane and interfère with time consuming, and problems related to environmental the catal}rtic reaction. At this stage we can only speculate contamination are common, so that constant monitoring is required (9). as to the nature of thèse substances, but ions such as nitrite In addition to thèse technical problems, there are other and ferrous iron have been shown to act as reducing agents reasons why measurement of UI is not widespread. The in the Sandell-Kolthoflf reaction (15). Whatever their na• validity of using Follis's criteria of categorizing casual ture, thèse other dialyzable interfering substances are urine samples on the basis of iodine/creatinine ratios (4) is responsible for up to 50% of the apparent increase in UI debatable, and Bourdoux has highlighted a number of when measured by automated dialysis in samples with UI disadvantages in using this index, especially in populations <100 /[ig/L. This is highlighted in Table 2, which shows the with lower body mass or protein intake (9). In addition, decrease in UI measured in automated dialysis, after the goitrogenic factors may exacerbate the iodine deficiency, samples had been ashed to remove any interfering sub• and measurement of UI excrétion alone will not always stances. For example, in the second sample the UI concen• reflect the severity of the eflfects of iodine deficiency in the tration decreased from 61 to 20 tigfh; therefore, 41 fj-g/L of population (19). the original UI concentration measured may have been due Using UI to investigate the efifects of iodine deficiency on to interfering substances. The USCN concentration in this a population is compounded by the fact that such effects are case was 5 mg/L, which produces an apparent iodine eflFect related primarily to a lack of maternai and hence fetal of 21.8 /xg/L in automated dialysis (see Figure 2). There• thyroid hormone (20). The déterminations of sérum thyroid fore, USCN accoimts for only 53% of the overestimation in hormones and thyrotropin are more-sensitive indices of the UI, whereas other substances may account for 47%. detrimental effects of iodine deficiency than is measure• Another important finding is the greater degree of over• ment of UI (21). Two new developments in measuring estimation or error due to the measurement of UI in thyrotropin have made this analyte more attractive for dialysis in samples with low iodine concentrations (<100 routine surveillance of an iodine-deficient population. /og/L). Using the régression équation generated in Figure 3, First, use of a small spot of blood dried on filter paper avoids one can calculate an expected 55.5 /Ag/L diflFerence between the collection of sérum, making this approach generally UI measured by dialysis and that by ashing for a sample more acceptable (22). Second, the précision and sensitivity with USCN of 10 mg/L. If this hypothetical sample had a of thyrotropin measurements have been vastly improved true UI concentration of 50 /xg/L, the overestimation of UI by the use of isotopic- and nonisotopic-labeled monoclonal would be 111%. However, if this same sample had a true UI antibody methods (23, 24). concentration of 300 jog/L, the overestimation is reduced to Clearly, the biochemical évaluation of the iodine status 18.5%. This accounts for the poor corrélation observed of communities and its application to the monitoring of when ail UI concentrations obtained by automated dialysis iodine prophylaxis require refinement. are plotted against USCN (r = 0.148, P = 0.388), because the degree of overestimation becomes less noticeable in samples with normal to high UI concentrations (>100 We gratefuUy acknowledge the financial support provided by the Australian International Development Assistance Bureau of the /ig/L). It may also explain the better corrélation observed Australian Government. This paper arose eut of the China- between digestion and dialysis by Gary et al. (5), whose Australia Technical Coopération Project on Iodine Deficiency. Our range of analyzed UI was apparently 100-1000 Mg/L. thanks also to Mrs. Krys O'Brien for her secretarial assistance. Our study draws attention to the serious problems of using the automated dialysis method for determining UI; Références in its current form, this method should not be used as a 1. Hetzel ES. Iodine deficiency disorders (IDD) and their eradica- monitoring tool for iodine deficiency. To our knowledge, tion. Lancet 1983;ii:1126-9. there is no other fully automated system for UI analysis 2. Boyages SC, Collins JK, Maberly GF, Jupp JJ, Morris JG, Eeistman CJ. Iodine deficiency impairs intellectual and neuromo- commercially available. The automated digestion system tor development in apparently normal persons. Med J Aust used in method B is more accurate, but lacks the simplicity 1989;150:676-9. and speed of dialysis and nécessitâtes the use of corrosive 3. Hetzel BS. Progress in the prévention and control of iodine concentrated minerai acids. Furthermore, the equipment deficiency disorders [Letter]. Lancet 1987;ii:266. used in method B is no longer available, having been 4. FoUis RH Jr. Pattems of urinary iodine excrétion in goitrous superseded by the dialysis system. Therefore, the available and non-goitrous areas. Am J Clin Nutr 1964;14:253-68. 5. Gary PJ, Lashley DW, Owen GM. Automated measurement of options for monitoring UI in a large number of samples are urinary iodine. Clin Chem 1973;19:950-3. limited. One possible alternative may be the use of a 6. Officiai methods of analysis of the Association of OflBcial Ana- semi-automated technique, but this has the disadvantage lytical Chemists, 13th ed. Arlington, VA: AOAC, 1980: Section 7, of requiring manual digestion of samples before the final 115. automated colorimetric analysis (16, 17). Because this 7. Belling GB. Further studies on the recovery of iodine as preliminary digestion uses either acid digestion or alkaline iodine-125 after alkaline ashing prier to assay. Analyst 1983;108:763-5. ashing, the sample-processing capacity of a laboratory is 8. Potter BJ, Jones GB, Buckley RA, Belling GB, Mcintosh GH, limited. Furthermore, the safety issues in using a concen• Hetzel HS. Production of severe iodine deficiency in sheep using a trated acid médium for sample digestion cannot be overem- prepared low iodine diet. Aust J Biol Sci 1980;33:53-61.

868 CLINICAL CHEMISTRY, Vol. 36, No. 6,1990 9. Bourdoux PP. Measurement of iodine in the assessment of mination of total iodine in food. Analyst 1980;105:344-52. iodine deficiency. IDD Newsl 1988;4:8-12. 18. Crowley LV, Jensen DR. An évaluation of an automated 10. Maberly GF. The aetiology, treatment and prévention of System for détermination of protein-bound iodine [Tech Brief]. Clin endémie goitre in Sarawak, Malaysia [MD Thesis]. Sydney: Uni- Chim Acta 1965;12:473-t. versity of New South Wales, 1983:119-33. 19. Delange F, Thilly C, Bourdoux P, Hennart P, Courtois P, 11. Bourdoux P, Delange F, Gerrard M, Mafuta M, Hanson A, Ermans AM. Influence of dietary goitrogens during pregnaney in Ermans AM. Evidence that cassava ingestion increases thiocy- humans on thyroid function of the newbom. Op. cit. (réf. i3):40-50. anate formation: a possible etiologic factor in endémie goitre. J 20. Vulsma T, Gons MH, De Vijler JJM. Matemal-fetal transfer of Clin Endocrinol Metab 1978;46:613-21. thyroxine in congénital hypothyroidism due to total organifîeation 12. Aldridge WN. The estimation of microquantities of eyanide defect or thyroid agenesis. N Engl J Med 1989;321:13-6. and thiocyanate. Analyst 1945;70:474-5. 21. Delange F. Iodine nutrition and congénital hypothyroidism. 13. Bourdoux P, Putzeys G, Lagasse R, Van Steirteghem A. Biochemical and statistical methods. In: Delange F, Iteke FB, In: Delange F, Fisher DA, Glinder D, Dussault JH, Ermans AM, Ermans AM, eds. Nutritional factors involved in the goitrogenic Irie M, eds. Research in congénital hypothyroidism. New York and aetion of cassava. Ottawa: International Development Research London: Plénum Press, 1989:173-85. Centre, 1982;IDRC-184e:20-4. 22. Waite KV, Maberly GF, Eastman CJ. Storage conditions and 14. Aumont G, Tressol JC. Improved routine method for the stability of thyrotropin and thyroid hormones on filter paper. Clin détermination of total iodine in urine and milk. Analyst Chem 1987;33:853-5. 1986;111:841-3. 23. Waite KV, Maberly GF, Ma G, Eastman CJ. Immunoradio- 15. Sandell EB, Kolthofif IM. Microdetermination of iodine by a metric assay with use of magnetizable particles: measurement of eatalytic method. Mikrochim Acta 1937;1:9-25. thyrotropin in blood spots, to screen for neonatal hypothyroidism. 16. Benotti J, Benotti N, Pino S, Gardyna H. Détermination of Clin Chem 1986;32:1966-8. total iodine in urine, stool, diets and tissue. Clin Chem 1965; 24. Torresani TE, Scherz R. Thyroid screening of neonates with- 11:932-6. out use of radioactivity: évaluation of time-resolved fluoroimmu- 17. Moxon RED, Dixon £J. Semi-automatic method for the déter• noassay of thyrotropin. Clin Chem 1986;32:10ia-5.

CLINICAL CHEMISTRY, Vol. 36, No. 6,1990 869