Resistance to Antibiotics With High Empiric Treatment Relevance Explains Attributable Mortality Across 110 Pathogen-antibiotic Combinations

To The Editor—We agree with Lee and Chen that drivers of AMR-associated mortality remain poorly understood and that many avenues of research remain under-explored. Our study found that antibiotic resistance among patients with bacteremia was associated with a 10% relative increase in mortality [1]; a small increase compared with many published estimates, including those used to measure the global burden of AMR [2]. We believe that the principal reason for this disparity is that most studies lack the ability to comprehensively adjust for patient healthcare exposures, comorbidities, and co-resistance patterns. We nevertheless identified a stronger relative increase (18%) for the subset of pathogen-antibiotic pairs deemed to have high empiric treatment relevance, based on a blinded adjudication by 2 study authors. Regarding Lee and Chen's first point, indeed resistance to antibiotics with low treatment relevance had zero direct impact on risk in our meta-regression model, which controlled for comorbidities and co-resistance patterns. But, if resistance to an antibiotic with low treatment relevance was associated with resistance to an antibiotic with high treatment relevance, then, indeed, this low treatment relevance resistance could have prognostic value. Regarding their second point, the consistently high estimates in Pseudomonas aeruginosa and Acinetobacter spp., were at least partly attributable to the fact that the antibiotics routinely reported for these organisms almost all had moderate-to-high empiric treatment relevance. Compare this to, say, Staphylococcus aureus, for which several routinely reported antibiotics, are of low empiric treatment relevance (eg erythromycin). As such, some of the organism-specific findings may reflect the wider principle of empiric treatment relevance. Regarding ceftriaxone resistance among Escherichia coli, Klebsiella spp., and Enterobacter spp., we dug in to extract the underlying hazard ratios and confidence intervals in Supplement 1.3 (see Table 1). The underlying data suggested little evidence of heterogeneity between the 3 hazard ratios in the main meta-regression (HR = 1.19, 95% CI: 0.90–1.56, I2 = 0%). However, heterogeneity was observed after adjustment for covariates but prior to adjustment for co-resistance (HR = 1.47, 95% CI: 1.17, 1.85, I2 = 65%). These results point to the difficulty in studying the drivers of mortality in persons with AMR infections—in addition to very strong confounding due to healthcare exposures and comorbidities, co-resistance patterns make it even more difficult, and noisy, to disentangle. Unadjusted and Adjusted Associations Between Ceftriaxone Resistance and 30-Day Mortality in Escherichia coli, Klebsiella spp., and Enterobacter spp. Extracted From Supplement 1.3 Abbreviation: HR, heart rate. Unadjusted and Adjusted Associations Between Ceftriaxone Resistance and 30-Day Mortality in Escherichia coli, Klebsiella spp., and Enterobacter spp. Extracted From Supplement 1.3 Abbreviation: HR, heart rate. Regarding their fourth point, our funnel plot is simply a testament to the large statistical variation in estimates across the 110 pathogen-antibiotic pairs. Very few of the estimates were precise, even in this multi-year study from a large jurisdiction. Finally, we agree that treatment norms do vary, and the authors raise an important point. In other jurisdictions, we may anticipate a somewhat distinct assessment of empiric treatment relevance. However, treatment relevance may still be associated with outcomes in a similar way, since adequacy of empiric therapy is such a strong predictor of survival [3]. Overall, our study found comparatively small impacts of AMR on bacteremia outcomes overall, but stronger impacts for antibiotics with high empiric treatment relevance. Further research and data to better understand the underlying drivers of outcomes among patients with bacteremia are needed. Financial support. The original study was funded by the Canadian Institutes for Health Research (grant number 401316).

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.