To the Editor,

With great interest, we read the recent original article of Matsuo et al.1 on detecting the signals of drug-drug interactions (DDIs) that increase the incidence of long QT syndrome. Matsuo et al. used 3,958 reports of long QT syndrome from all reports; 611,336 reports in the Japanese Adverse Drug Event database (JADER).1

Matsuo et al. calculated the reporting odds ratios (RORs) and 95% CIs of the concomitant use group and single drug use groups separately as described above. Based on previous reports,2, 3 Matsuo et al. predicted that the possibility of an adverse event caused by a suspected DDI was expected to be increased if the ROR of the coadministration group was higher than those of single-use groups and these 95% CIs were mutually exclusive.

However, we recently reported that the ROR scores based on the frequency-based statistics (definition by Susuta et al.4) are easily inflated; thus, the use of the “upward variation of ROR scores” to search for DDI signals increases the likelihood of false-positive signal detection.5 This means that the validity of the DDI signal reported by Matsuo et al. is unclear.

Many algorithms have also been reported to search for DDI signals.6, 7 Of them, the Ω shrinkage measure,8, 9 proposed by Norén et al., is used by the World Health Organization Uppsala Monitoring Center (WHO-UMC), and previous study has shown that it has the most conservative signal detection trend among signal detection methods based on frequentist statistics.10 Therefore, we calculated the Ω shrinkage measure scores and 95% CIs from the raw data in Table 3 of article by Matsuo et al (See the Appendix for the calculation method).

Of the 31 signal-positive combinations with 33 drugs reported by Matsuo et al., 4 signal-negative combinations are Amiodarone & Metildigoxin (Ω025: −0.15), Donepezil & Famotidine (Ω025: −0.25), Donepezil & Solifenacin (Ω025: −0.38) and Famotidine & Digoxin (Ω025: −0.17) (Table S1). Amiodarone and Metildigoxin had risk rating D (modify regimen), however, based on Digoxin in Table 4. Furthermore, the remaining 3 combinations had risk rating of A (no interaction) by Lexicomp®.1 Based on these results, it is highly likely that these 4 combinations were false-positive signals. In this study, the possible false-positive signal is at least 4/31 (12.9%), which is not a lot.

Although there is currently no definitive method for detecting signals of DDIs, the active use of algorithms utilizing ROR would not be recommended due to the existence of the Ω shrinkage measure, which indicates a conservative detection trend. Given the importance of safety signals obtained from spontaneous reporting systems, researchers are advisable to select detection algorithms that are likely to have fewer false-positive signals.

All authors declare that there are no competing interests.