Optimisation of pyrogen testing in parenterals according to different pharmacopoeias by probabilistic modelling

Sebastian Hoffmann, Ursula Lüderitz-Püchel, Thomas Montag, Thomas Hartung

Research output: Contribution to journalArticlepeer-review


The rabbit test to detect pyrogenic contamination in parenterals is crucial to ensure patient safety. The pharmacopoeial tests in Europe, the US and Japan are based on the fever reaction of rabbits, but differ in their experimental design and in their algorithms to assess contamination. Employing an international reference endotoxin, fever can be induced in rabbits. Data from 171 rabbits built the base for probabilistic modelling of the fever reaction and for the comparison of the pharmacopoeial tests. The rabbit fever reaction could be modelled as a function of the amount of injected endotoxin (per kg body weight) by linear regression. Combining the pharmacopoeial algorithms of the rabbit pyrogen test with the developed model allowed analysis of differences regarding test results and animal consumption. This showed that the assessment of pyrogenic contamination strongly depends on the respective pyrogen test stipulated by regulations. Additionally, the approach was used to develop a new experimental design. Two specific versions of this design resulted in a reduction of the number of animals used by about 30% while the safety of the test was maintained. A need for harmnonisation is evident, allowing optimisation of the experimental design, which promotes animal welfare.

Original languageEnglish (US)
Pages (from-to)25-31
Number of pages7
JournalJournal of Endotoxin Research
Issue number1
StatePublished - 2005
Externally publishedYes


  • Animal model
  • Animal welfare
  • Pharmacopoeias
  • Pyrogens
  • Statistical model

ASJC Scopus subject areas

  • Microbiology
  • Immunology
  • Molecular Biology
  • Cell Biology
  • Infectious Diseases


Dive into the research topics of 'Optimisation of pyrogen testing in parenterals according to different pharmacopoeias by probabilistic modelling'. Together they form a unique fingerprint.

Cite this