Myelosuppression with Oxazolidinones: Are There Differences?

LETTER

We read with interest Lee & Caffrey’s paper (1) comparing thrombocytopenia events in association with tedizolid and linezolid from the FDA Adverse Event Reporting System (FAERS), a database of voluntary spontaneous AE reports submitted to the FDA (https://www.fda.gov/drugs/guidancecomplianceregulatoryinformation/surveillance/adversedrugeffects/default.Htm). We fully agree with their conclusion that “…thrombocytopenia with tedizolid should be monitored…,” since it is well established that oxazolidinones can be associated with myelosuppression, especially during long-term treatment (>2 weeks) (2, 3). However, we would like to note three points about their analyses.

First, Lee & Caffrey identified thrombocytopenia in just a single report of a tedizolid-treated patient. Due to this small number, the resultant reporting odds and proportional reporting ratios had extremely wide 95% confidence intervals, and caution should be exercised when interpreting such results.

Second, FAERS data analysis has well-known limitations, which is clearly acknowledged by the FDA (https://www.fda.gov/drugs/guidancecomplianceregulatoryinformation/surveillance/adversedrugeffects/default.Htm); such analyses are considered to be hypothesis generating and not a means for evaluating safety signals (i.e., confirming causal associations between drugs and AEs) (4). A notable limitation of the FAERS database is the lack of sufficient information on all potential confounding factors. For instance, it is unknown if this patient had received other potentially myelosuppressive drugs (including linezolid), either prior to or concomitantly with tedizolid. Without such information, causal associations are difficult to make. Reporting rates following a product’s first approval also have to be interpreted cautiously in light of noncausal factors influencing spontaneous reporting (5). Physicians tend to report AEs for unanticipated/unpublished safety issues more frequently in the first 5 years postapproval and may be more vigilant with new products regarding specific safety concerns known from other members of the same class (the “Weber Effect”) (6). New antibacterials, like tedizolid, also tend to be reserved for sicker and more complicated patients immediately postapproval (“channeling bias”), which confounds safety comparisons between drugs that are at different time points postapproval (e.g., 18 and 4 years for linezolid and tedizolid, respectively) (7).

Third, considering those data imperfections, statistical methods that can account for the “noise” inherent in spontaneous reports to safety databases must be used. Disproportionality analyses (as performed by Lee & Caffrey), which do not have denominators of actual patient exposure or total number of observed events during drug exposure, cannot be used to calculate incidence or estimate risks/relative risks accurately. Health authorities use better statistical methods (e.g., multi-item Gamma Poisson Shrinker) to describe observed-to-expected event rates in complex safety databases (4, 5).

Given these points, Lee & Caffrey’s analyses appear to be insufficient to draw conclusions about the comparative safety profiles between tedizolid and linezolid, which requires randomized controlled clinical trials with appropriate monitoring of laboratory parameters. To date, tedizolid and linezolid were compared in one phase 1 trial (both drugs for 21 days), three phase 3 trials evaluating 6 days of tedizolid versus 10 days of linezolid, and one phase 3 trial evaluating both drugs for 7 to 21 days (8–12). In all of these trials, which assessed relatively short treatment durations, tedizolid resulted in similar or lower rates of adverse hematologic outcomes and/or drug-related AEs relevant to myelosuppression compared to linezolid. We encourage prospective long-term studies comparing myelosuppression risks between oxazolidinones in order to rigorously evaluate potential differences (or a lack thereof).