Introduction:
The interest in safety assessments of herbal products is increasing, but not their quality, as limitations of available data are often not sufficiently taken into account. To adress the issue, a classification of sources of bias and ways out is aimed to.
Materials and Methods:
A systematic data base search for reviews in this field, combined with hand searching in text books, was conducted. Sources of bias were classified according to data types involved.
Results and Discussion:
Depending from the data involved, different sources of bias were identified:
- Data on quality: Often the great differences of the composition of herbal products prepared from the same plant are neglected, so leading to flaws when tranferring data.
- Non-clinical data: Common pitfalls are the transfer of data from in vitro studies to the clinical setting, without taking into account the influence of ADME. Often also effects from sublethal high-dose settings are used without sufficiently taking into account dose dependency or, especially e.g. in carcinogenicity studies, methodological ambiguities [1].
- Data from clinical studies and post marketing surveillance: Lack of differentiation between negative studies and failed studies leads to wrong conclusions on inefficacy, the evaluation of safety data is often flawed by neglecting background incidences as e.g. in case of hepatotoxicity [2], by protopathic bias, and by the awareness and views of authors of case reports [3].
Conclusions:
A higher awareness of common pitfalls in the assessment of safety data on herbal products is needed, e.g. in case of hepatotoxic risks, if we want to avoid that methodological artefacts and misperceptions of the generalizability of data continue to influence our view of the safety of herbal products, both by neglecting risks, as, more abundant, by exaggerating non-existing risks.
References:
[1] IARC 2016, IARC monographs, vol.108; [2] Andrade et al., Pharmacogenomics. 2009 10:1467-87; [3] Horwitz and Feinstein 1980, Am J Med. 68:255-8