On 2014 Nov 28, Juergen Borlak commented:

There has already been an extensive communication on this matter (see comments on the original manuscript in PLOS One at http://www.plosone.org/annotation/listThread.action?root=69409). Unfortunately, we note serious misconceptions that require clarification.

First, in our previous communications we already alluded to the methodology employed. Thus, all relevant information can be retrieved from the original paper. Furthermore, in our original study we alerted to the issue of underreporting and the difficulties in defining accurate incidences of ADRs. Apart from statistical issues, e.g. to determine an incidence of just 1 per 10,000 in the control group one would require 180,000 participants in the verum group, prospective studies are inevitably confounded by the study protocol inclusion and exclusion criteria that lead to additional bias. Second, in 2013 the pharmacovigilance risk assessment committee (PRAC) of the European Medical Agency (EMA) evaluated independently the evidence of flupirtine induced liver injury and concluded that the benefit still outweighs the risk for hepatotoxicity particularly when the treatment of acute pain with other analgesics (e.g. non-steroidal anti-inflammatory drugs, weak opioids) is contraindicated and the treatment duration is restricted to 14 days (http://www.ema.europa.eu/docs/en_GB/document_library/Referrals_document/Flupirtine-containing_medicines/Recommendation_provided_by_Pharmacovigilance_Risk_Assessment_Committee/WC500146103.pdf). In our previous communication we referred to the deliberations of the PRAC, however, based on the comment provided we must assume that the commentator has no confidence in the capabilities of the PRAC and this would also imply that the EMA apparently is unaware of the issue of ADR underreporting which is a somewhat baffling assumption. Lastly, we do not believe that an estimate of incidences of ADR without causality assessment adds to the science of drug safety. Indeed, an incidence of less than 1 in 10,000 may be found acceptable; however, for an individual suffering from adverse drug reactions the incidence is intolerable. Therefore, we encourage the commentator to focus on the main thesis of our study, that is to improve pharmacovigilance through causality assessment and to identify individuals at risk for ADR prior to medication, as was recently reported by us for paracetamol (see Borlak et al., Genome Medicine 2013, 5(9):86).

On 2014 Nov 08, Dirk Wetzel commented:

Basic epidemiological terms should be used correctly

I respectfully disagree with the methods and conclusions of this publication. The article suggests that an incidence of flupirtine-related hepatobiliary adverse events of “1:100,000 …or 0.8 in 10,000” has been calculated. In fact, this "incidence" is based on reported cases only. The estimated number of unreported cases might be as high as 90 to 95%, because considerable under-reporting is a well-known phenomenon of spontaneous reporting. Incidence rate is clearly defined as the number of new cases of a certain condition in relation to the population at risk within a specified period of time. As the authors used the number of reported cases as numerator (“new cases”), this method probably leads to a dramatic under-estimation of the potential risks associated with flupirtine. This could be interpreted as downplaying the risks of flupirtine.

On 2014 Nov 08, Dirk Wetzel commented:

Basic epidemiological terms should be used correctly

I respectfully disagree with the methods and conclusions of this publication. The article suggests that an incidence of flupirtine-related hepatobiliary adverse events of “1:100,000 …or 0.8 in 10,000” has been calculated. In fact, this "incidence" is based on reported cases only. The estimated number of unreported cases might be as high as 90 to 95%, because considerable under-reporting is a well-known phenomenon of spontaneous reporting. Incidence rate is clearly defined as the number of new cases of a certain condition in relation to the population at risk within a specified period of time. As the authors used the number of reported cases as numerator (“new cases”), this method probably leads to a dramatic under-estimation of the potential risks associated with flupirtine. This could be interpreted as downplaying the risks of flupirtine.