und 400 cm- 1 bei einem Datenpunktabstand von 4 cm- ~ ausge- wertet. Lediglich der Bereich zwischen 2500 und 1800 cm- wird ausgeblendet, da hier keine relevante Information enthal- ten ist.

Das skizzierte Verfahren wurde bis heute anhand einer Reihe von Natursteinen getestet. Es zeigte sich, dab eine Analyse der Harnsteine in einer geniigenden Genauigkeit mrglich ist. Das Auswerteverfahren ist bisher auf die Bestimmung der hS, ufigsten Harnsteinkomponenten beschr/inkt. Trotzdem werden in der Routineanalytik die iiberwiegende Anzahl der F/ille erfaBt.

Dutch die Weiterentwicklung der hier zur Diskussion gestell- ten Methode soil erreicht werden, dab eine automatische infra- rot-spektroskopische Harnsteinanalyse routinem/il3ig mrglich wird.

Literatur

1. Hesse A, Bach D (1982) Harnsteine - Pathobiochemie und klinisch-chemische Diagnostik. Thieme, Stuttgart

2. Hesse A, Molt K (1982) J Clin Chem Clin Biochem 20:861 - 873

3. Hesse A, Sanders G (1988) Infrarotspektren - Atlas zur Harnsteinanalyse. Thieme, Stuttgart New York

4. Malinowski ER, Howery DG (1980) Factor analysis in che- mistry. Wiley, New York

5. Frederics PM, Lee JB, Osborn PR, Swinkels DAJ (1985) Appl Spectrosc 39:303, 311

F r e s e n i u s Z A n a l C h e m ( 1 9 8 8 ) 3 3 0 : 3 7 2 3 7 3 �9 S p r i n g e r - V e r l a g 1 9 8 8

A 32 On-line result interpretation of eerebrospinal fluid protein proffies with the Behring Nephelometer BNA M. Wick ~, A. Fateh-Moghadam ~, Ch. Przetak ~, K. Einhiiupl ~, M. Knedel ~, P. Tiingler 3, and M. Niih 3

Institute for Clinical Chemistry and 2 Department of Neurology at Klinikum GroBhadern, Ludwig-Maximilians-Universitfit, Marchioninistrasse 15, D-8000 Miinchen 70, Federal Republic of Germany 3 Behringwerke A.G., D-3550 Marburg, Federal Republic of Germany

On-line Interpretation yon Proteinprofilen der Cerebrospinalfliissigkeit mit dem Behring Nephdometer B N A

For diagnosis of central nervous system diseases, traditional methods for clinico-chemical analysis of CSF proteins such as colloidal reactions or conventional electrophoretic techniques are being replaced increasingly by immunoassays of single pro- teins as IgG and albumin and high resolution electrophoretic methods, as f.i. isoelectric focusing.

Since the concentrations of CSF proteins depend on their plasma levels, their hydrodynamic radii as well as on blood CSF

barrier permeability, CSF turnover and a possible local synthesis within the CNS, the interpretation of results requires an appro- priate model taking into consideration all these influences.

To meet this requirement even under high demand for CSF protein profiles we developed a computer assisted method for standardized result presentation and interpretation of IgG and albumin determinations performed by the Behring Nephel- ometer Analyzer (Fateh-Moghadam). The interpretation is based on the concepts of Felgenhauer und Reiber [1-3] , employing the original Reiber's diagramme [2], which we de- veloped further according to our own empirical experience (Fig. 1) in order to introduce graduations in the assessment of both barrier dysfunctions and CNS immune responses [4]. Furthermore, the adequate processing of borderline result pat- terns required the insertion of an equivalent of a "fuzzy logic" employed in knowledge based systems, which was solved by introducing transitional sectors based on empirical experience taking into consideration both results of concomitant analyses as f.i. oligoclonal IgG and the imprecision of methods used as well as the biological variance and the age of the patient. Owing to the age-dependency of the blood-CSF-barrier function, mar- ginal CSF/serum albumin ratios have to be dealt with accord- ingly. For this purpose the CSF/serum albumin ratio was con- sidered to be normal (including a safety margin for the impre- cision of the method) up to 5 x 10- 3 for children of 1 to 15 years of age, up to 7.5 x 10 -3 for the group 1 5 - 4 0 years and up

Fig. 1 Reiber's diagramme, modified for computer-assisted result interpretation especially in marginal result patterns. IgG syn = [(IgG CSF - IgG Serum/ 369)-(Alb. CSF - Alb. Serum/230) (IgG Serum/Alb. Serum) x 0.43] x 0.5 [g/24 h] (according to

Tourtellotte). IgG syn = (IgG CSF/IgG Serum - 0.8 • ]/(Alb. CSF/Alb. Serum) 2 + 15 + 1.8) x IgG Serum [g/l] (according to Reiber)

IgG CSF x 1 0 0 0 IgG Serum ~ ~

:~ �9 .,..v..' ~,,.:~.,:~ :,:.,2 ~:.." ~::~ .,. ~..,.: ~.~ .~ ~.:: ~ ::: ':~:':,, '.~..'.~::~..~ f,.'.'.':::|

x<7,4 ~ 7,4<x <_ x)23 �9 ~4.:,.*..::~:,,'~'~"#~:~,: ~':

�9 -,.~ . ~ .~ �9 " " ~ 17

N~.~:" ~*,." ~ ' N 11 ~ ' ~ ~::::":~:~ ::"~I ..:~ ~ ....... o

N -~."-~ :'@ .... N

5 ~:" '1 ~

1 A= AIb. CSF x 1 0 0 0 ~ ~ Serum . . . . .

) 0 ~ 10 15 20 25 3'0

373

os te 5 A

to 10 • 10 -3 for adults beyond 40 years. However, since the immaturity of the blood-CSF-barrier and of the immune system in newborn infants and, concomitantly, the vast biological vari- ance attributable to different degrees of maturity at birth pre- clude a reliable interpretation of CSF protein profiles, the appli- cation of the system in neonatology was excluded (safety mar- gin: up to 1 year of age).

To each sector of the adapted diagramme an interpretation is assigned, which is supplemented in ambiguous cases by re- commendations for further investigations suited to the solution of the diagnostic problem, as oligoclonal IgG.

Approximate quantitation of IgG synthesis rate according to Tourtellotte's [5] or Reiber's [3] formulas can yield sup- plementary information.

The serum IgG and albumin concentrations are assessed additionally with regard to a possible increase or decrease.

Result interpretations issued by the system were compared to their manually produced counterparts in 229 unselected patients. In 97 % a complete concordance of both interpretations was observed, whereas gradual deviations found in 3% were predominantly attributable to borderline result patterns.

Furthermore, the clinical validity of interpretations was con- firmed in 336 unselected patients with ascertained CNS diseases.

We introduce a new, easy to handle, fully mechanized and computer assisted method for result presentation and interpre- tation of CSF protein profiles offering a few advantages:

1. Quick and synoptic presentation of results, 2. Automatic calculation of CSF/serum IgG and albumin

ratios as well as CNS IgG synthesis rate, 3. Standardized result interpretation suitable for disease

monitoring, 4. Separate assessment of serum results, 5. Low demand of labour, time and cost, 6. Interpretations with regard to the age and the tentative

diagnosis of the patient.

However, the following limitations have to be taken into consideration:

1. System dependent on analytical accuracy of results, es- pecially the avoidance of high-dose-hook-effect,

2. System dependent on correct taking of samples, especially - equilibrium of CSF and plasma compartment - no traumatic spinal tap,

3. System not applicable in neonatology (up to I year of age) owing to the immaturity of blood-CSF-barrier and immune system,

4. System in its present version valid only for CSF taken by lumbar puncture, for cisternal CSF a correction factor has to be introduced. For ventricular CSF, however, in most cases involving a ventricular shunt the equilibrium between CSF and plasma as a essential precondition cannot be guaranteed.

References

1. Felgenhauer K, Schliep G, Rapic N (1976) J Neurol Sci 30:113-128

2. Reiber H (1980) J Neurol 224:89-99 3. Reiber H, Felgenhauer K (1987) Immunoglobulin Cerebro-

spinal Fluid (CSF)/Serum quotients as a function of blood CSF barrier permeability. Quantitation of local immun- globulin synthesis in the central nervous system. In: Metab- olism and development of the nervous system. Wiley, Chichester, UK

4. Fateh-Moghadam A, Przetak Ch, Wick M, Einhfiupl K, Knedel M (in preparation)

5. Tourtelotte WW, Potvin AR, Fleming JK, Murthy KN, Levy J, Syndulko K, Potvin JH (1980) Neurology 30:240-244

Fresenius Z Anal Chem (1988) 330:373-374 �9 Springer-Verlag 1988

A 33 Validit/it des Sekretin-Pankreozymin-Tests (SPT) zur Diagnostik der exkretorischen Pankreasinsuffizienz - Ergebnisse einer Leaving-one-out-Studie Walter Dummler 1, Eva Siegmund t und Friedrich Hauzeur 2

1 Institut ffir Pathologische Biochemie der Wilhelm-Pieck-Universitfit Rostock, Ernst-Heydemann-Strasse 6, DDR-2500 Rostock, Deutsche Demokratische Republik 2 Klinik ffir Innere Medizin des Bezirkskrankenhauses Rostock, Otto-Grotewohl-Ring, DDR 2500-Rostock, Deutsche Demokratische Republik

Diagnostic validity of the secretin-CCK test in the diagnosis of exocrine pancreatic insufficiency - results of a leaving-one-out-study

Zur Diagnostik einer Pankreasinsuffizienz ist der SPT die aus- sageffihigste Methode. Wenn zur Bewertung der Testergebnisse Referenzbereiche herangezogcn werden, liegen Sensitivitfit und SpezifitM jedoch bei h6chstens 80%. Die Effizienz lfil3t sich durch multivariate Auswertung wesentlich verbessern [1]. Unser Ziel war es, den Merkmalsvektor zu optimieren und die Gtite des Tests mit Hilfe des Leaving-one-out-Verfahrens zu prfifen.

Methodik

1. Lernstichprobe. Sic bestand aus 45 Patienten mit gesicherter chronischer Pankreatitis (Klasse K: mittleres Lebensalter =

374

40,3 Jahre, mittlere Erkrankungsdauer = 5,0 Jahre) und 45 Pa- tienten ohne Pankreasschaden (Klasse G: 23 yon ihnen mit ,,funktionellen Oberbauchbeschwerden"; mittleres Lebensal- ter = 38,3 Jahre).

2. SP-Test. Nach Gewinnung yon Duodenalsaft fiber 2 x 10 rain (Basalsekret) erhielten die Probanden je 1 U/kg K6rpergewicht Sekretin und Pankreozymin intraven6s. Danach wurde der Duodenalsaft tiber 4 x 10 rain gesammelt (Fraktionen 1 - 4) und innerhalb 30 min analysiert. Auger dem Volumen wurden in allen Fraktionen die Hydrogencarbonatkonzentration und die Aktivitfiten von Trypsin (mit BAPNA), Lipase (mit Olivenrl), Amylase (mit Amylose) und Chymotrypsin (mit SUPHEPA) gemessen. Somit standen je Proband neben Lebensalter und Krrpergewicht 6 x 6 primfire und 5 x 6 sekund~re Ergebnisse (Outputs) als Diskriminanzmerkmale zur Verf/igung.

3. Auswertung. Sic erfolgte mit einem multiplen linearen Regres- sionsmodell, dessen Vorteil darin besteht, dag es Normalvertei- lung der Originaldaten und Varianzhomogenitgt nicht voraus- setzt.

Die Diskriminanzfunktion ftir einenp-dimensionalen Merk- malsvektor und ein Objekt j lautet:

bo = ~ s s = xij N, N = ArK + NG �9

i = 1 j

Die Reklassifizierung der Objekte erfolgte anhand der Werte der Diskriminanzfunktion d i sowie aufgrund der Wahrschein- lichkeit P(K)~ der Zugeh6rigkeit eines Probanden zur Klasse