Faecal Nitrogen Determination by Near-Infrared ...

Eur J Clin Chem Clin Biochem 1996; 34:561 -563

© 1996 by Walter de Gruyter · Berlin · New York

SHORT COMMUNICATION

Faecal Nitrogen Determination by Near-Infrared Spectroscopy Otto Bekers\ Coble Postma1, Johan C. Fischet*, Paul K H. Franck1 and Arnold J. P. F. Lombard 1 2

Department of Clinical Chemistry and Haematology, Leyenburg Hospital, The Hague, The Netherlands Department of Clinical Chemistry, Wilhelmina Children's Hospital, Utrecht University, Utrecht, The Netherlands

Summary: The well-known Kjeldahl method for the determination of faecal nitrogen is rather complex, time-consuming and expensive. Therefore, the use of near-infrared spectroscopy in determining the amount of nitrogen in faeces has been studied. To our knowledge we are the first to present the calibration equation for the determination of nitrogen with near-infrared spectroscopy. A good correlation (r = 0.96) was found between results from near-infrared spectroscopy and the Kjeldahl method. The imprecision of both methods was comparable. Once the rather laborious calibration has been performed, near-infrared spectroscopy is shown to be a very simple and rapid method for measuring nitrogen in faeces. Introduction Measurement of nitrogen in faeces is important for the detection and evaluation of malabsorption of nitrogenous nutrients and for performing energy balance studies, especially in patients with cystic fibrosis (1—4). The method used to quantify nitrogen in faeces in clinical laboratories, is the well-known Kjeldahl procedure (5). This method is rather complex, tedious and time-consuming, making it less attractive for serial analyses (2). The use of aggressive chemicals is another disadvantage. An alternative method might be near-infrared spectroscopy, a technique increasingly used in clinical chemistry (4, 6-8). Quite recently Picarelli et al. (4) described the use of nearinfrared spectroscopy for the quantitative determination of faecal fat, nitrogen and water. However, the article from Picarelli presents no data on the calibration procedures of the near-infrared spectroscopy method. In an earlier report we described the application of near-infrared spectroscopy for the determination of faecal fat (8). The present study confirms that near-infrared spectroscopy can also be used advantageously for the analysis of nitrogen in faeces. Materials and Methods Eighty-four patients (children, n = 34; adult cystic fibrosis patients, n = 32; miscellaneous internal patients, n = 18) whose clinical picture suggested malabsorption were studied. Fifty-five of them were used for construction of the calibration curve and the remaining 29 patients for the comparison of the methods. Determination of faecal nitrogen was performed on stools from one-day or three-day collections. The faeces of the three-day collections were put together before analysing. After weighing, stools were homogenised and the nitrogen content was determined by the Kjedahl method (5) as well as by nearinfrared spectroscopy (8). The near-infrared spectroscopic analyses were carried out on an InfraAJyser® Model 400 (Bran and Luebbe, Maarssen, The Netherlands) connected to an IBM personal computer PS2 Model 55SX. The software APC® was supplied by Bran and Luebbe (9). After

stool homogenisation, about 5 grams of the sample were removed with a spatula and placed on a disposable cup (Bran and Luebbe) and smoothed flat with a glass slide. The prepared sample cup was placed in an interchangeable sample drawer and then pushed into the InfraAlyser®. The result is given within one minute. Regression analyses for method comparisons were done by using the Passing & Bablok procedure (10, 11). Results and Discussion Calibration of the near-infrared spectroscopy method To obtain reliable results of nitrogen in faeces with near-infrared spectroscopy the InfraAlyser® first had to be calibrated (9). To this end the nitrogen concentrations of samples were determined with the Kjeldahl method. Δ-Alanine was used for the assessment of the accuracy of the Kjeldahl method. For the calibration, 55 samples were used, randomly chosen among the three patients groups and with nitrogen concentrations distributed over the entire range of interest (0.2—1.5 mol/kg). The distribution over this range was: 0.2-0.6 mol/kg: 11 samples; 0.6-1.0 mol/kg: 29 samples; 1.01.5 mol/kg: 15 samples. These samples were scanned by near-infrared spectroscopy in the region from 1100 nm to 2500 nm, the reflectances of the samples being detected by the InfraAlyser®. The wavelengths and scaling factors (F), that were most closely aligned to the nitrogen content analysed with the Kjeldahl method, were selected. This calibration process is described in more detail in our earlier report (8). The complex multivariate regression analyses (9, 12) were performed by the software program APC®, which is supplied with the InfraAlyser®. The results of this study are presented in table 1. The values from table 1 can be substituted in the following equation (Eq. 1): (Eq. 1)

Nitrogen = Z + Fn log (l/Rn)

where Z is the bias correction, Fn the scaling factor for wavelength n and Rn the reflectance at wavelength n. This results in the calibration equation (Eq. 2): Tab. 1 The essential parameter results of multivariate regression analyses for the calibration of faecal nitrogen by near-infrared spectroscopy found in a study of 55 samples (cf. Eq. 1). Wavelength (nm)

Scaling factor

1680 1759 1778 2100 2190 2348 Za

607.1 2746 -4053 -193.5 1229 -341.6 71.60

α

Ζ is the bias correction

Short communication

562 Nitrogen = 71.60 + + + -

607.1 log 1/R|680 2746 log l/Ri?59 4053 log l/Ri778 193.5 log l/R 2 ioo 1229 log l/R 2 i90 341.61og*l/R2348

Tab. 2 Imprecision of the faecal nitrogen assays by the Kjeldahl procedure and the near-infrared method. Kjeldahl

(Eq. 2)

We constructed a calibration curve since hitherto no data concerning the calibration of nitrogen in faeces by near-infrared spectroscopy has been reported. Nevertheless, it is advisable to check this calibration equation before use, since as we suggested in our paper on the determination of faecal fat, calibration equations may be dependent on matrix effects of stools due to differences in food composition in various countries (8). Correlation study The relationship between the faecal nitrogen content in 29 samples analysed both with the Kjeldahl method and with near-infrared spectroscopy, using the calibration curve as described in equation 2, is shown in figure 1. The correlation between both methods, calculated as described by Passing & Bablok (10, 11), is y = 0.91 χ + 0.09 (y = near-infrared spectroscopy; χ = Kjeldahl). The correlation coefficient (r) was found to be 0.96 and the standard error of the estimate (SEE) was 0.05 mol/kg. These data are somewhat better than those described by Picarelli et al. (4), who found a coefficient of correlation of 0.92. Imprecision of the results Table 2 summarises the results of the intra-day and inter-day imprecision. The results of the inter-day imprecision were obtained within a period of two weeks. The samples were stored at 4 °C. The imprecision of both methods is seen to be of the same order of magnitude. Stability of the samples Samples can be stored at -20 °C for at least four months. Eight samples with nitrogen concentrations over the entire range of interest were analysed weekly during a four months period. During this

n xa (mol/kg) SDb (mol/kg) CV° (%) a b c

Sample number

a

Intra-day

Inter-day

Intra-day

Inter-day

12 1.03 0.03

13 1.03 0.06

9 0.95 0.03

10 1.16 0.04 3.4

3.2

5.8

2.9

5L = mean value of n determinations SD = standard deviation CV = coefficient of variation

Tab. 3

1 2 3 4 5 6

Near-infrared method

Stability of samples after four freezing-thawing cycles. Faecal nitrogen content (mol/kg)

CV3

Freezing-thawing cycle number

0

1

0.43 0.56 0.83 0.95 1.00 1.21

0.41 0.53 0.84 0.92 1.02 1.17

0.39 0.57 0.79 0.97 1.03 1.21

0.43 0.57 0.80 0.94 0.99 1.24

0.42 0.56 0.81 0.94 0.97 1.26

4.0 3.0 2.5 1.9 2.4 2.8

CV = coefficient of variation

period no degradation of the nitrogen content was observed; all coefficients of variation were below 5%, which is comparable with the inter-day coefficient of variation (see tab. 2). A sample can be thawed and afterwards frozen again at least four times without influencing the outcome of the result. For this study six samples were analysed; the results are presented in table 3. No degradation was observed and all coefficients of variation were below 4%.

1.4

Quality control Since we have shown that faeces samples stored at -20 °C are stable for at least four months, these samples are suitable for the short-term quality control of the near-infrared method. Long term quality control must be performed by periodically analysing samples using both near-infrared spectroscopy and the Kjeldahl method. Conclusions

0.0

0.7

1.4

Faecal nitrogen (Kjeldahl method) [mol/kg)

Fig. 1 Correlation between faecal nitrogen results (mol/kg) (n = 29), measured by near-infrared spectroscopy and the Kjeldahl procedure; y = 0.91 χ + 0.09; r = 0.96; p < 0.01.

Near-infrared spectroscopy offers the following benefits to the Kjeldahl method: the analysis-time (about 1 minute per sample) is short; the method is simple to perform and the use of (aggressive) reagents is obviated. The imprecision of both methods is comparable and there is a good correlation between results from nearinfrared spectroscopy and the Kjeldahl method. On the other hand, the technique has to be performed using a sophisticated instrument which brings with it a high price. Therefore, the use of near-infrared spectroscopy seems most suitable in specialised or research centres with a substantial amount of samples to be analysed.

References 1. Eastham RD. Biochemical values in clinical medicine. Bristol: Wright, 1985:241-2.

2. Walmsley RN, White GH. A guide to diagnostic clinical chemistry. Chicago: Blackwell Scientific Publication, 1988:326-7.

Short communication 3. Murphy JL, Wootton S A, Bond S A, Jackson AA. Energy content of stools in normal healthy controls and patients with cystic fibrosis. Arch Dis Child 1991; 66:495-500. 4. Picarelli A, Greco M, Di Giovambattista F, Ramazzotti A, Cedrone C, Corazziari E, et al. Quantitative determination of faecal fat, nitrogen and water by means of a spectrophotometric technique: near infrared reflectance analysis (NIRA). Assessment of its accuracy and reproducibility compared with chemical methods. Clin Chim Acta 1995; 234:147-56. 5. Archibald RM. Nitrogen by the Kjeldahl method. In: Seligson D, editor. Standard methods of clinical chemistry. New York: Academic Press, 1958; 2:91-9. 6. Hall JH, Pollard A. Near-infrared spectroscopy: a new dimension in clinical chemistry. Clin Chem 1992; 38:1623-31. 7. Hall JH, Pollard A. Near-infrared spectroscopic determination of serum total proteins, albumin, globulins and urea. Clin Biochem 1993; 26:483-90. 8. Bekers O, Postma C, Lombarts AJPF. Determination of faecal fat by near-infrared spectroscopy. Eur J Clin Chem Clin Biochem 1995; 33:83-6.

563 9. Alfa-Level group, Bran and Luebbe. Reference manual analyzing technologies. Publication no. TL8-4546-10. Theory of near-infrared reflectance analysis. New York: Elmsford, 1987; Al:l-32. 10. Passing H, Bablok W. A new biometrical procedure for testing the equality of measurements from two different analytical methods. J Clin Chem Clin Biochem 1983; 21:709-20. 11. Passing H, Bablok W. Comparison of several regression procedures for method comparison studies and determination of sample sizes. J Clin Chem Clin Biochem 1984; 22:431-45. 12. Beebe K, Kowalski BR. An introduction to multivariate calibration and analysis. Anal Chem 1987; 59:1007-17. Received Febmaiy 7/April 29. 1996 Corresponding author: Dr. O. Bekers, Leyenburg Hospital, Department of Clinical Chemistry and Haematology, P. O. Box 40.551, NL-2504 LN The Hague, The Netherlands

Eur J Clin Chem Clin Biochem 1996; 34:565-568 © 1996 by Walter de Gruyter · Berlin · New York

SHORT COMMUNICATION

Population-Based Differences in Thyrotropin and Thyroxine Distributions in Healthy Newborns Revealing Results from Independent Reagent Evaluation Tamara Tuwninen1, Gennady Tsukerman2 and Jean-Louis Dhondf* 1 2 3

Labsystems Research Laboratories, Helsinki, Finland Byelorussian Institute for Hereditary Diseases, Minsk, Republic of Byelarus Centre Regional de Depistage, Faculte de Medicine, Lille, France

Summary: An independent evaluation of reagents for the determination of thyroxine and thyrotropin from dried blood spot samples taken from newborns between the third and fifth day of life revealed striking differences in the thyrotropin distribution among newborns from Byelorussia. An analysis of the thyrotropin distribution from Byelorussian newboms showed that 40% of samples had over 5 mlU/1 blood. In other European populations comparable in respect to timing of blood collection, this fraction varied from only 1% (Stockholm, Sweden) to 3.7% (Lille, France). The reason for the "shift right" in Byelorussian newborns remains to be further investigated. This shift can be attributed to the synergetic effects of mild-moderate iodine deficiency and/or as yet unidentified environmental factors.

this porportion is the most variable in screening programs. Similar age distribution was assumed for the Florida study.

The differences observed in cumulative distribution patterns obtained by two commercial methods question the use of absolute figures (such as the proportion of samples over 5 mIU/1) for the purpose of inter-population comparisons.

In Europe, thyrotropin screening is used as a primary strategy. Samples over the cut-off are tested and/or recalled; the decision is usually made on the basis of thyrotropin concentration increase. In Minsk, Moscow and Stockholm, total thyroxine or free thyroxine is not routinely determined from samples with thyrotropin increase, but in Lille, a pilot programme on the determination of free thyroxine from samples with elevated thyrotropin was initiated.

Introduction It is well-documented that in screening for congenital hypothyroidism, the time of sample collection (1), exposure of the mother to iodine contrasting agents and antithyroid drugs (2, 3) or iodine disinfectants (4, 5), transplacental transfer of thyrotropin-receptor blocking antibodies (6), and the course of labour can affect the levels of thyroxine, and to a greater extent, the levels of thyrotropin. It is also an established fact that even mild endemic iodine deficiency stimulates an increased secretion of thyrotropin which serves as a good marker in population-based screening programs for the eradication of micronutrient malnutrition (7). The effect of industrial pollution on thyroid status has been also reported (8). In this communication we describe the findings of differences in thyroxine and thyrotropin distributions in newborns when commercial reagents were independently evaluated in different geographical areas. The reasons for these differences are discussed below. Materials and Methods Samples Blood samples in the Minsk, Moscow, Stockholm and Lille areas were collected from the third to the fifth day of life according to accepted practices. In the USA, the current practice is to collect samples generally before 48 hours of age (1). Contrary to the European cohorts regarding age, the US cohorts were heterogeneous. In the Virginia study, for example, the age distribution was the following: 16% in the age group 0-24 h after birth, 52% 24-48 h, 10% 48-72 h, 4% 72h-l week and more than 1 week-17%. A high proportion of samples over 1 week represented recall samples;

In Minsk, Stockholm, Lille, Virginia and Florida, samples were analysed shortly after their arrival to the laboratory. Only the samples from Moscow were shipped to Helsinki (Labsystems Research Laboratories) and analysed after 5 months of storage at — 20 °C. Supported by literature reports (9) and our observations (unpublished data), we assumed that at these storage conditions the anaiytes were stable, and that the results from the Moscow cohort were comparable to the respective results from other cohorts. Screening algorithms and cut-off decision

Due to the early discharge from the nurseries in the USA, thyroxine is used as a primary test in the majority of states. Samples below the daily 10th percentile are retested for thyrotropin determination. According to the stipulations of the Federal Food and Drug Administration, those samples showing thyrotropin above the 90th percentile are reported as abnormal. For comparison, the data from the Virginia study was excluded because thyrotropin was determined only from the samples with thyroxine values below the 10th percentile, thus a priori thyrotropin values shifted to the right. Laboratories and quality control All screening laboratories involved in this study operate according to good laboratory practice and participate regularly in national and/or international quality assurance programmes for performance evaluation and quality control. Labsystems Research Laboratories is an ISO-9001 certified institution which also participates in European (Deutsche Gesellschaft für Klinische Chemie, Bonn, Germany) and US (Centers for Disease Control and Prevention, Atlanta, Georgia) Quality Assurance Programs. Reagents, their characteristics, assay performance and statistics Neonatal hTSH FEIA and Neonatal T4 FEIA (Labsystems, Helsinki, Finland) were used for evaluations according to the kit instructions. All European samples were analysed with calibrators prepared on Schleicher & Schuell paper grade 2992 (Dassel, Germany), whereas US samples were analysed with calibrators prepared on Schleicher & Schuell paper grade 903 (USA). Each of

Short communication

566 Tab. 1 Assay performance characteristics

Low detection limit

Neonatal hTSH FEIA (mIU/I blood)

Neonatal T4 FEIA (nmol/l blood)

1

12

Value

CV (%)

Value

CV (%)

Within-run imprecision

17a 36 71

11.2 8.8 6.9

28b 44 86

11:2

Between-run imprecision

16a 33 60

9.1 8.9 10.9

28b 47 92

11.2 7.8 6.5

a b

5.0 6.5

over proposed cut-off presumably below the cut-off

the evaluations was performed blind with regard to the others. Measures of frequency distribution were calculated and compared after all evaluations were completed. To illustrate the dependency of thyrotropin frequency distributions on commercial reagents, we compared respective distributions in two European newborn cohorts to the respective results by Delfia TSH IFMA (Wallac, Turku, Finland). The cohorts were from iodine sufficient (10, 11) and ecologically clean areas (Stockholm and Lille). Cumulative thyrotropin frequency distributions were calculated for both methods. The summary of the assay performance characteristics is presented in table 1. Results The measures of frequency distribution of thyrotropin and thyroxine in different cohorts are presented in table 2. For consistency, all values refer to the volume of whole blood. Table 2 shows that the thyrotropin distributions in Moscow, Stockholm and Lille are fairly similar, with a very low proportion of samples above 5 mIU/1 blood, ranging from 1 % (Stockholm) to 3.7% (Lille). The distribution of thyrotropin in Florida is shifted to the right, and the proportion of samples over 5 mlU/1 blood exceeded 20%. This shift in the distribution is explained by the high proportion of samples taken before 72 h after birth and represented

Tab. 2

Table 3 shows that thyroxine distribution patterns were similar in both USA testing sites, with a greater proportion of higher values due to physiological thyroxine surge which is, however, not as evident as the thyrotropin surge. On the contrary, compared to the thyroxine distribution in Lille, the respective thyroxine distributions in Moscow and Minsk cohorts showed a shift to the left. The distribution in Lille seems to be in agreement with the reference range for thyroxine for 1 -4 day-old neonates, calculated as a 95% normal interval being 66-146 nmol/l blood ,(12). Dependency of thyrotropin frequency distributions on commercial reagents is shown in figures 1 and 2. These figures show cumulative thyrotropin frequency obtained with two commercial methods in iodine replete (10, 11) and ecologically clean areas. As seen from figure 1, the thyrotropin cumulative distributions in Stockholm are steep, and the proportion of samples over 5 mIU/1 blood shown by both methods is fairly comparable. The difference in the distribution pattern by the two methods was more evident as the distributions became wider (fig. 2). Thus, the proportion of samples over 5 mIU/1 blood was 3.7% in Method A and exceeded 11% in Method B.

Discussion With the current Neonatal hTSH FEIA reagent, in Moscow, Lille and Stockholm areas, a cut-off of 10 mIU/1 blood could be recommended (see frequency distribution, tab. 2). In the USA, according to the stipulation of the Federal Food and Drug Administration, the cut-off for thyrotropin should be based on the 90th percentile, equivalent to 8 mIU/1 blood. A conservative strategy of using the 90th percentile as a cut-off for reporting abnormal samples can eventually result in high recall rates. To reduce recall rates, ageadjusted cut-offs for thyrotropin would probably be an appropriate solution. Newborn, rather than school-aged, thyrotropin screening was reported to be a more sensitive indicator of a community iodine status (8). The proportion of babies with thyrotropin values exceeding 5 mIU/1 serum as well as recall rates in Australia, Canada and South Poland were compared in population-based studies to demonstrate overload with industrial pollutants in South Poland (8). The calculation of the sample proportion above 5 mIU/1 as a criterion for assessing population thyroid status was proposed by the Centers for Disease Control and Prevention. This was adopted in

Comparison of the frequency distribution of thyrotropin (mIU/1 blood) Minsk (Byelorussia) n = 1544

Median 95th percentile 99th percentile 99.9th percentile Proportion (%) of samples, > 5 mJU/1

Tab. 3

the physiological thyrotropin surge. Distribution of thyrotropin in the Minsk newborn cohort is even more shifted to the right with the respective fraction of samples over 5 mIU/1 blood reaching 40%.

4.0 22 34 36

=«40

Moscow (Russia) n = 878

Stockholm (Sweden) n = 3098

0.5 4.9 6.9 8.4 2.4

0.5 1.9 5.1 8.4 1

Lille (France) n = 3936

0.5 4.5 8.0 12.3

3.7

Florida (USA) n = 517

1.8 8.9 21 27 20.6

Comparison of the frequency distribution of thyroxine (nmol/l blood)

10th percentile 20th percentile Median

Minsk (Byelorussia) n = 2684

Moscow (Russia) n = 725

Lille (France) n = 1390

Florida (USA) n = 681

Virginia (USA) n = 398

20 32 57

33 38 52

45 57 70

55 68 86

52 60 98

Short communication

567

9

11

10

12

13

14

15

16

17

18

—f— 19

-H 20

Thyrotropin [mlU/l]

Fig. 1 Comparison of thyrotropin cumulative frequency distribution (n = 3098) obtained by two commercial methods in the iodine

replete and ecologically clean area of Stockholm (Sweden). The solid line represents Method A and the dotted line — Method B.

The cumulative frequency displayed in figure 1 is as follows: Thyrotropin (mlU/l blood)

Method A (%)

Method B (%)

Thyrotropin (mIU/1 blood)

Method A