Reply to “Comment on ‘Gram-positive targeting antibiotics are associated with progression and death in women with platinum-sensitive recurrent high-grade epithelial ovarian cancer’”

Dear Editor,

We thank Dr. Zhu for the thoughtful commentary regarding our study evaluating the association between gram-positive targeting antibiotics and oncologic outcomes among women receiving platinum-based chemotherapy for platinum-sensitive recurrent ovarian cancer [1,2]. We agree that residual confounding, particularly confounding related to underlying comorbidity, infection severity, and patient frailty, is an important consideration when interpreting these findings.

We wish to take this opportunity to provide additional clarity regarding one aspect of the commentary. Our multivariable models adjusted for stage and residual disease at the time of initial cytoreductive surgery; they did not adjust for body mass index or diabetes [1]. We explicitly identified the higher prevalence of diabetes and increased body mass index among patients receiving antibiotics as important potential confounders [1]. We further acknowledged that our retrospective data could not determine whether the observed outcomes were attributable to gram-positive targeting antibiotic exposure itself, the physiologic consequences of infection, increased susceptibility to infection among patients with greater medical frailty, or a combination of these factors. Therefore, our conclusions (and title) were intentionally framed as associations rather than evidence of causality [1].

Several findings provide some reassurance that the association cannot be entirely explained by differences in baseline ovarian cancer prognosis. The study groups had comparable Eastern Cooperative Oncology Group (ECOG) performance status and no significant differences in stage, residual disease, hereditary cancer mutations, secondary cytoreductive surgery, or receipt of bevacizumab, poly(ADP-ribose) polymerase inhibitors, or hyperthermic intraperitoneal chemotherapy (HIPEC) [1]. In addition, time to first progression, which preceded the antibiotic exposure period evaluated in this study, was similar among the antibiotic exposure groups [1]. Thus, the oncologic management and survival trajectories of both the treatment groups appear balanced prior to antibiotic intervention. Nevertheless, these observations do not eliminate the residual confounding the authors highlight in their letter. The small number of patients exposed to gram-positive targeting antibiotics, 20 patients, representing 7.7% of the cohort, also limited the number of covariates that could be reliably incorporated into multivariable models. Infection severity, hospitalization, corticosteroid exposure, antimicrobial indication, microbiologic findings, and other measures of comorbidity or physiologic reserve should be captured prospectively in future studies.

Although residual confounding remains possible, the present findings are consistent with our prior clinical and preclinical observations [3,4]. In a retrospective cohort of 424 women with newly diagnosed stage III–IV epithelial ovarian cancer, antibiotic exposure during platinum chemotherapy was associated with shorter progression-free and overall survival, with the strongest associations observed among patients receiving antibiotics targeting gram-positive organisms [3]. Complementary mechanistic studies demonstrated that broad-spectrum antibiotic-mediated disruption of the gut microbiota accelerated ovarian tumor growth and reduced cisplatin sensitivity in murine models. Importantly, restoration of the microbiota through cecal transplantation from untreated mice ameliorated chemoresistance and prolonged survival, supporting a potentially microbiome-mediated effect [4]. Together, these findings across front-line and recurrent clinical settings, supported by preclinical evidence, strengthen the biological plausibility of an association between antibiotic-mediated microbiome disruption and platinum response. However, we wish to remain clear that they do not establish causality or demonstrate that any specific antibiotic class directly worsens outcomes in patients.

We also agree with Dr. Zhu regarding the importance of antimicrobial stewardship. However, we would caution against using our findings to preferentially select empiric “gram-negative” therapy or to withhold clinically indicated gram-positive coverage. Our study did not compare antimicrobial treatment strategies, evaluate the adequacy of infection management, or assess infection-related outcomes. It therefore cannot determine whether one empiric regimen is safer or more effective than another in patients receiving ovarian cancer therapy. For high-risk patients with febrile neutropenia, guideline-concordant initial therapy generally consists of antipseudomonal β-lactam monotherapy, such as cefepime, piperacillin-tazobactam, meropenem, or imipenem-cilastatin [5,6]. These agents are broad-spectrum antibiotics and are more accurately characterized as antipseudomonal β-lactams than as “gram-negative regimens.” The routine addition of vancomycin or another agent targeting resistant gram-positive organisms is not recommended as part of the standard initial regimen [5,6]. Additional gram-positive coverage should be reserved for defined clinical circumstances, including hemodynamic instability, suspected catheter-related infection, skin or soft-tissue infection, pneumonia with concern for methicillin-resistant Staphylococcus aureus, microbiologic evidence of a resistant gram-positive pathogen, or other patient- and institution-specific resistance risks [5,6].

Our findings should not be used to delay or withhold clinically indicated treatment of a potentially life-threatening infection. The study did not compare empiric antibiotic regimens, assess the appropriateness of infection management, or evaluate infection-related morbidity and mortality. It therefore cannot establish that avoidance of gram-positive coverage improves outcomes. Rather, the findings reinforce established stewardship principles: obtain appropriate cultures when feasible, select therapy according to the suspected source and clinical severity, avoid unnecessary anti-resistant gram-positive coverage, reassess treatment as clinical and microbiologic information becomes available, and de-escalate promptly when continued coverage is not supported [5,6].

Our findings should instead be viewed as hypothesis-generating evidence supporting prospective investigation of how antibiotic indication, spectrum, timing, and duration influence the microbiome and treatment response during platinum chemotherapy. Future studies incorporating longitudinal microbiome sampling, microbial functional analyses, infection-related variables, comorbidity measures, and antimicrobial exposure will be necessary to distinguish a potential microbiome-mediated effect from confounding by infection and baseline patient vulnerability [7].

We again thank Dr. Zhu for highlighting these important considerations and for advancing discussion regarding antimicrobial stewardship and microbiome-informed cancer care.