Antibiotic-induced myasthenia worsening—an estimation of risk based on reporting frequency

Abstract

Sir, A number of drugs, including antibiotics, are notorious for causing sometimes rapid deterioration in patients with myasthenia.1 To guide the clinician’s choice there are lists of substances that can interfere with neuromuscular transmission.2–4 However, some of the substances in such a list may have been included because of only a few reports. Moreover, a substance not listed could indeed be safe or excluded because of an arbitrary threshold of reports in the literature. Ideally, the clinician would appreciate a white list of drugs explicitly labelled as carrying a low risk of worsening myasthenia. As an alternative to lists of substances to avoid, the relative risk of all antimicrobial options may be estimated from pharmacovigilance databases. For the statins, this approach was pursued recently.5 However, that study addressed whether statins per se carry a risk of worsening myasthenia. Thus, all drugs other than statins were used as a reference in that study. In contrast, we compared the myasthenia-worsening risk only within the group of antibiotics. Therefore, we used all other antibiotic drugs from a predefined list as a reference to calculate the reporting OR (ROR) for a given antibiotic. ROR is the OR derived from a 2 × 2 contingency table as illustrated in Table S1 (available as Supplementary data at JAC Online). If RORD0<1⁠, a particular drug is less likely to be reported as causing worsening of myasthenia than all other antibiotics. Whether this difference was significant or not was assessed with the standard χ2 test (1 degree of freedom), with χ2 = 3.841 corresponding to P = 0.05.8 Antibiotics with ≥5000 ADRs reported in VigiBase® n.s., not significant; *significantly (P < 0.05) reduced below average; **significantly (P < 0.05) increased above average. Number of ADRs in connection with a myasthenic reaction (fourth column), number of all other ADRs reported (fifth column), ROR (sixth column), χ12 (seventh column), as derived from the 2 × 2 contingency table that underlies the calculation of ROR, and significance level (eighth column) of χ2 test (1 degree of freedom). 0.26 = ROR value as calculated by Gras-Champel et al.5 153 = ROR value as calculated by Gras-Champel et al.5 2.26 = ROR value as calculated by Gras-Champel et al.5 As a test of the sensitivity of the reporting retrieval process, we also retrieved ADR counts for paracetamol (used as a zero control), penicillamine (used as positive control) and atorvastatin (RORs all reported by Gras-Champel et al.5). To calculate the RORs of these non-antibiotic control substances, a and c were calculated as described above, whereas for the calculation of b and d the sum extended across all antibiotic drugs of our list (but did not include the two remaining non-antibiotic control drugs). Table 1 summarizes a=nD0, Myasthenia⁠, c=nD0, not Myasthenia and RORD0⁠. An ROR significantly <1 (which we will refer to as ‘low risk’) was found for many penicillins, cephalosporins, co-trimoxazole and rifampicin. A higher-than-average ROR (ROR >1, ‘high risk’) was calculated for most macrolides, fluoroquinolones and two of four aminoglycosides. For the remaining antibiotics, RORs did not differ significantly from 1 (which we will refer to as ‘intermediate risk’). For some of these (e.g. fosfomycin), the total number of reported events is small, leading to insignificant χ2 despite low RORs. In others (e.g. vancomycin), there is a high number of total events and the ROR is close to 1, indicating an average risk probably higher than in the low-risk group. The quantitative analysis confirmed the clinical impression of a high risk of aminoglycosides, macrolides and fluoroquinolones.1 Interestingly, tetracyclines were not associated with frequent reporting, although there are in vitro data that imply a higher risk.2,9 Low RORs for cephalosporins and penicillins are in accordance with clinical experience. Ampicillin, in contrast to its low ROR value, has sometimes been listed as requiring some caution.2 A low ROR was also found for combinations of penicillins with β-lactamase inhibitors. Although there are reports of myasthenia patients deteriorating with vancomycin,10 the database showed that the reporting frequency indicated intermediate risk. Therefore, if the underlying infectious agent requires the use of vancomycin the risk seems acceptable. For substances introduced more recently (e.g. ceftobiprole or ceftolozane) the total number of events reported is too small to calculate reliable values for χ2. Our figures should be interpreted with caution. ADRs of different antibiotics could be reported with different probability. If clinicians were aware of a particular complication, they may be more likely to report that event. In patients at risk of worsening of myasthenia, clinicians may avoid antimicrobials perceived to be of high risk, resulting in low numbers of reports for these drugs. Since the individual reports are not accessible, concomitant medication could also be responsible for an ADR. The entries in Table 1 are a snapshot as ADRs are continuously reported to the database. From 27 February 2020 (first retrieval of data) to 25 May 2020 (retrieval of data for the final analysis) we noted an increase by 39 myasthenia-related ADRs and by 53 873 ADRs not related to myasthenia (from a baseline of 1 900 000) for the drugs included in a previous version of our study. For amoxicillin (seven new myasthenia-related ADRs) and for metronidazole (five new myasthenia-related ADRs) the ROR changed from significantly below 1 to average. Nonetheless, our figures reproduced known patterns of high and low risk options. They help to rank additional important antibiotics that are not commonly mentioned in lists as low risk. From the risk estimation introduced here, co-trimoxazole and rifampicin may be grouped as low risk (comparable to cephalosporins/penicillins). Vancomycin and meropenem are ranked as intermediate risk according to our calculation. The study was carried out as part of our routine work. None to declare. Table S1 is available as Supplementary data at JAC Online.

Topics & Concepts