DÄ internationalArchive39/2009Requirements and Assessment of Laboratory Tests—Part 5 of a Series on Evaluation of Scientific Publications: In reply

Correspondence

Requirements and Assessment of Laboratory Tests—Part 5 of a Series on Evaluation of Scientific Publications: In reply

Dtsch Arztebl Int 2009; 106(39): 638; DOI: 10.3238/arztebl.2009.0638

Bautsch, W

I can only agree with Professor Hopf. The indication to conduct a test for borreliosis seems uncontroversial for the “hard” indicators that he lists. Statements that the borreliosis test may be helpful in unclear clinical presentations (I assume that these are the so called „soft“ indicators), as occasionally heard, should, however, be considered as highly problematic. In this scenario, the available serological tests are inappropriate, in my opinion, and should therefore not even be ordered.

Seroprevalence data can of course be included in estimating the positive predictive value (and would lower this value in such a scenario). However, such data are scarce in Germany; their geographical distribution is heterogeneous, and they change over time (after all, borreliosis seems to be on the increase). Serodiagnostics can yield more information than just “positive”/”negative”: the class of antibodies (IgM/IgG), the species of borrelia, and band pattern on the immunoblot test show a certain degree of correlation to the clinical signs of borreliosis, so that I am currently not able to provide a generally valid, quantitative answer here.

In response to Eising’s correspondence, I wish to clarify that my position as expressed in my article is by no means one of general opposition to screening tests. Neonatal screening or screening blood specimens for markers of infection are often useful or even unavoidable. In my article I wanted to draw attention to a general problem that applies to any investigation: the higher the number of healthy subjects in the study, the higher the number of false positive results. Without having more precise details about the specificity and sensitivity of the testing procedure and the prevalence of the disease that is being tested for, it is impossible to assess rationally the clinical value of suggested screening measures. This also holds true for the suggested inpatient computed tomography screening of patients older than 70, of course..

Whether “serious comorbidities” need to be ruled out rapidly in inpatients strongly depends on the definition of comorbidities. Naturally, this goes without saying if these are associated with the underlying disease (for example, cytomegalovirus in AIDS), as their prevalence in the study cohort would be high. However, it would undoubtedly be highly controversial to conduct extensive investigations to rule out tumors in a 20 year old female patient with appendicitis without corresponding symptoms—especially as the question would then arise why such investigations are performed only in patients and not in the entire population.

I very much doubt whether it will become possible in the near future to distinguish genuine positive results from false positive results by using computer aided evaluation of the results of a laboratory screening test, independently of what information is available on the prevalence of possible pathologies. I am not aware of any good examples of such an approach. The only thing that springs to mind in this context are artificial neural networks (1,2), although as far as I am aware these have thus far been used only in symptomatic patients.
DOI: 10.3238/arztebl.2009.0638


Prof. Dr. med. habil. Dr. rer. nat. Wilfried Bautsch
Institut für Mikrobiologie, Immunologie
und Krankenhaushygiene
Celler Straße 38
38814 Braunschweig
w.bautsch@klinikum-braunschweig.de

Conflict of interest statement
The authors of all contributions declare that no conflict of interest exists according to the guidelines of the International Committee of Medical Journal Editors
1.
Astin ML, Wilding P: Application of neural networks to the interpretation of laboratory data in cancer diagnosis. Clin Chem 1992; 38: 34–8. MEDLINE
2.
Schwarzer G, Schumacher M: Artificial neural networks for diagnosis and prognosis in prostate cancer. Semin. Urol Oncol 2002; 20: 89–95. MEDLINE
3.
Bautsch W: Requirements and assessment of laboratory tests—part 5 of a series on evaluation of scientific publications [Anforderungen und Bewertung der Ergebnisse von Laboruntersuchungen]. Dtsch Arztebl Int 2009; 106: 403–6. VOLLTEXT
1. Astin ML, Wilding P: Application of neural networks to the interpretation of laboratory data in cancer diagnosis. Clin Chem 1992; 38: 34–8. MEDLINE
2. Schwarzer G, Schumacher M: Artificial neural networks for diagnosis and prognosis in prostate cancer. Semin. Urol Oncol 2002; 20: 89–95. MEDLINE
3. Bautsch W: Requirements and assessment of laboratory tests—part 5 of a series on evaluation of scientific publications [Anforderungen und Bewertung der Ergebnisse von Laboruntersuchungen]. Dtsch Arztebl Int 2009; 106: 403–6. VOLLTEXT