Antibiotics and probiotics differentially shape immunotherapy outcomes in non-small cell lung cancer

Antibiotic exposure and survival outcomes

The studies included in the antibiotic cohort were predominantly retrospective real-world investigations and enrolled patients with NSCLC receiving ICIs, most of whom had advanced disease. The antibiotic-related analysis was comprised of 34 studies that included patients treated with ICIs mainly in the second-line or later settings (Table 1). Across studies, immunotherapy regimens were largely based on PD-1/PD-L1 inhibitors. Antibiotic exposure was generally defined as systemic antibiotic use within predefined peri-ICI time windows and the antibiotics administered were mainly broad-spectrum agents. The meta-analysis demonstrated that antibiotic use was significantly associated with unfavorable survival outcomes in patients with NSCLC receiving immunotherapy. As shown in Figure 1A, antibiotic exposure was associated with a significantly reduced ORR [pooled odds ratio (OR) = 0.60, 95% confidence interval (CI) = 0.43–0.84, P = 0.002]. Antibiotic exposure was also associated with a significantly shorter OS [pooled hazard ratio (HR) = 1.68, 95% CI = 1.49–1.91, P < 0.001] and PFS (pooled HR = 1.62, 95% CI = 1.37–1.92, P < 0.001; Figure 1B, C). These findings are biologically plausible. Antibiotics can markedly reduce gut microbial diversity, disrupt the homeostasis of beneficial commensal bacteria, and potentially alter the pulmonary immune microenvironment via the gut-lung axis. Consistent with this notion, clinical studies have reported that antibiotic administration, particularly when given before or during immunotherapy, is closely associated with diminished ICI efficacy³⁶. Collectively, antibiotic exposure exhibited a trend toward a reduced ORR, while being consistently and significantly associated with an inferior OS and PFS, which indicated that the detrimental impact on immunotherapy outcomes in NSCLC is more robustly captured by long-term survival endpoints than by early response metrics. This finding also suggests that antibiotic-induced disruption of the gut microbiota may exert sustained and systemic inhibitory effects on antitumor immunity rather than merely delaying initial tumor control. In this context, antibiotic exposure during critical windows surrounding immunotherapy initiation may be particularly relevant, underscoring the need for careful consideration of antibiotic use in patients with NSCLC receiving ICIs³⁷.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Effects of antibiotics and probiotics on immunotherapy outcomes in patients with non–small cell lung cancer. (A–C) Antibiotics: ORR, OS, and PFS. Squares indicate study-specific effects; diamonds show pooled estimates with 95% CIs. (D–F) Probiotics: ORR, OS, and PFS. Squares indicate study-specific effects; diamonds show pooled estimates with 95% CIs. CI, confidence interval; HR, hazard ratio; ORR, objective response rate; OS, overall survival; PFS, progression-free survival.

Subgroup analyses by line of therapy showed a broadly consistent association between antibiotic exposure and worse outcomes. The association was more evident in first-line and mixed-line settings for the ORR with pooled ORs of 0.65 (95% CI = 0.48–0.87; I² = 1%) and 0.42 (95% CI = 0.25–0.70; I² = 56%), whereas the estimate in second-line and later studies was not significant and showed substantial heterogeneity (OR = 1.85, 95% CI = 0.66–5.18). However, no significant subgroup difference was observed (P = 0.103). Antibiotic exposure was associated with shorter OS across first-line, second-line and later therapies, and mixed-line populations with pooled HRs of 1.47 (95% CI = 1.19–1.82; I² = 91%), 2.27 (95% CI = 1.55–3.31; I² = 0%), and 2.00 (95% CI = 1.59–2.52; I² = 73%), respectively. PFS had a similar pattern with pooled HRs of 1.58 (95% CI = 1.04–2.39; I² = 82%), 1.33 (95% CI = 0.91–1.96; I² = 29%), and 1.83 (95% CI = 1.39–2.40; I² = 85%) in the corresponding subgroups. Neither OS nor PFS differed significantly across treatment-line strata (P = 0.173 and P = 0.648, respectively).

The results were similar when stratified by immune checkpoint blockade strategy. Antibiotic exposure remained associated with worse OS in anti-PD-1-only studies (HR = 1.56, 95% CI = 1.29–1.88; I² = 89%), anti-PD-L1-only studies (HR = 1.20, 95% CI = 1.04–1.39), anti-PD-1/anti-PD-L1 studies (HR = 1.76, 95% CI = 1.38–2.23; I² = 75%), and mixed-strategy studies (HR = 1.93, 95% CI = 1.48–2.51; I² = 47%) without a significant subgroup difference (P = 0.390). Pooled HRs for PFS were 2.00 (95% CI = 1.37–2.93; I² = 74%), 1.10 (95% CI = 0.97–1.25), 1.58 (95% CI = 1.22–2.05; I² = 68%), and 1.61 (95% CI = 1.09–2.37; I² = 76%) across these strata with no significant interaction (P = 0.536). The ORR was less consistent. Specifically, the reduction in response was most apparent in anti-PD-1/anti-PD-L1 studies (OR = 0.45, 95% CI = 0.23–0.88; I² = 62%), whereas the estimates in anti-PD-1-only (OR = 0.59, 95% CI = 0.23–1.48; I² = 70%) and mixed-strategy studies (OR = 0.77, 95% CI = 0.49–1.22; I² = 39%) were not significant; no subgroup difference was identified (P = 0.601). Overall, the adverse association of antibiotic exposure was more consistent for OS and PFS than the ORR.

Probiotic supplementation and therapeutic benefit

Probiotic-related evidence is comprised of 10 studies, most of which were retrospective cohort or real-world comparative analyses (Table S1). These studies encompassed patients receiving ICIs across first- to third-line treatment settings. The formulations varied across studies with respect to probiotic interventions but preparations based on Clostridium butyricum MIYAIRI 588 (CBM588) were most frequently used with Bifidobacterium spp. and mixed probiotic formulations. Probiotic interventions were typically administered orally as long-term adjuncts, like CBM588, to immunotherapy with control groups consisting of patients receiving identical immunotherapy regimens without probiotic supplementation. Probiotic supplementation was significantly associated with more favorable clinical outcomes in contrast to the unfavorable associations observed with antibiotics³⁸. Probiotic use was associated with a significantly higher ORR compared to no probiotic use (pooled OR = 1.84, 95% CI = 1.36–2.49, P < 0.001), as well as a significantly prolonged OS (pooled HR = 0.60, 95% CI = 0.46–0.79, P < 0.001) and PFS (pooled HR = 0.63, 95% CI = 0.52–0.76, P < 0.001), as illustrated in Figure 1D–F. Although the probiotic formulations and administration strategies varied across studies, the consistent improvement in the ORR and survival outcomes suggested that microbiota-supportive interventions (especially CBM588) may enhance antitumor treatment responses and improve long-term prognosis in NSCLC. These findings provided a rationale for considering microbiota modulation as an adjunctive strategy to optimize therapeutic outcomes in clinical practice.

Assessment of publication bias

Potential publication bias was assessed using funnel plots (Figure S1). Visual inspection of the funnel plots showed no obvious marked asymmetry in the probiotic analyses, whereas some asymmetry was observed in the antibiotic analyses, particularly for OS and PFS. Nevertheless, the number of studies was limited for several comparisons and these findings should be interpreted with caution. Further large-scale prospective studies are warranted to confirm the present results.

Sensitivity analyses

Sensitivity analyses using a leave-one-out approach showed that the pooled estimates remained largely unchanged after sequential exclusion of individual studies (Figure S2). The associations with OS and PFS were consistently maintained for antibiotic exposure with minimal variation in effect size and statistical significance. A similar pattern was observed for ORR, indicating that the overall findings were not disproportionately influenced by any single dataset.

Sensitivity analyses likewise demonstrated stable effect directions across all evaluated outcomes for probiotic supplementation. Despite the limited number of studies, the pooled estimates remained broadly consistent, supporting the relative robustness of the observed associations.

Discussion

This meta-analysis synthesized currently available real-world evidence to evaluate the impact of microbiota-modulating exposures on immunotherapy outcomes in patients with NSCLC. The results showed that antibiotic exposure is consistently associated with poorer survival outcomes and is unfavorably associated with a reduction in the ORR. In contrast, probiotic supplementation demonstrated a favorable trend in treatment outcomes, although the current evidence base remains limited. Collectively, these observations underscore the emerging importance of host-related determinants, particularly gut microbiota composition, in shaping the durability of benefit from ICI.

The differential strength of the associations observed for survival outcomes and objective response suggests that early tumor regression and long-term immunotherapy benefit may be driven by partially distinct biological mechanisms. Initial tumor response is influenced by tumor-intrinsic characteristics, including tumor immunogenicity and the baseline immune microenvironment, whereas durable clinical benefit depends on sustained activation and functional maintenance of antitumor immune responses. Antibiotic-associated disruption of gut microbial homeostasis may interfere with antigen presentation, impair dendritic cell maturation, and weaken memory T-cell responses, thereby contributing to both reduced treatment response and less durable immune-mediated tumor control³⁹.

Subgroup analyses indicated that the adverse impact of antibiotic exposure on immunotherapy outcomes was generally consistent across different lines of therapy and immune checkpoint inhibitor strategies with no significant differences observed between subgroups. This pattern suggests that the observed variability may partly reflect methodologic heterogeneity across studies rather than true biological differences. Because most included studies did not systematically report key antibiotic-related variables, such as antibiotic class, exposure timing, and indication for use, more refined stratified analyses could not be performed.

Current treatment recommendations for NSCLC are primarily based on disease stage, resectability, oncogenic driver status, and PD-L1 expression. Although ICIs have been incorporated into standard treatment algorithms across multiple clinical settings, current clinical guidelines do not provide explicit recommendations regarding antibiotic use or probiotic supplementation during immunotherapy, and microbiota-related factors are not included in routine treatment stratification.

NSCLC-specific evidence indicates that antibiotic exposure during immunotherapy is associated with clinically meaningful adverse outcomes. Peri-treatment antibiotic use has been linked to a significantly shorter PFS (pooled HR = 1.47, 95% CI = 1.13–1.90) and OS (pooled HR = 1.69, 95% CI = 1.25–2.29) in patients with advanced or metastatic NSCLC treated predominantly with ICI monotherapy, which corresponded to an absolute reduction of approximately 6–7 months in median survival. The negative association appears most pronounced when antibiotics are administered within the peri-immunotherapy window, supporting a potential interaction between early microbiota disruption and immune activation⁴⁰.

Similar trends have been reported in first-line chemo-immunotherapy (chemo-IO), which represents a current standard treatment paradigm for advanced NSCLC without actionable oncogenic drivers. Antibiotic exposure within 60 days before treatment initiation was associated with a lower ORR (27% vs. 40%), shorter PFS (3.9 vs. 5.9 months; HR = 1.35, 95% CI = 1.10–1.60), and reduced OS (10 vs. 15 months; HR = 1.50, 95% CI = 1.20–1.80) in a large multicenter cohort. Consistent findings have also been demonstrated in meta-analytic evaluations, demonstrating inferior survival in the overall IO-treated population (HR = 1.93, 95% CI = 1.52–2.45) and in patients receiving chemo-IO specifically (HR = 1.54, 95% CI = 1.28–1.84)⁴¹.

These observations suggested that the adverse impact of antibiotic exposure may be particularly relevant in advanced-stage disease, in which durable immune-mediated tumor control is critical. From a clinical perspective, a more cautious and indication-driven approach to antibiotic use during the peri-ICI period may therefore be warranted. In contrast, although microbiota modulation through probiotic supplementation has been proposed as a potential strategy to enhance immunotherapy efficacy, current evidence remains insufficient to support routine implementation in unselected patients.

Several limitations to this meta-analysis should be acknowledged. First, most included studies were retrospective observational analyses and causal inference remains limited. Second, the overall number of eligible studies, particularly for probiotic interventions, was relatively small, which may have affected the statistical power and stability of pooled estimates. Although probiotic formulations were broadly comparable in the immunomodulatory profiles, methodologic variability in dosing strategies, treatment duration, and study design may still introduce residual confounding. Third, heterogeneity across studies and the lack of detailed stratified data precluded more refined analyses of specific microbiota-related intervention strategies.

In conclusion, antibiotic exposure is associated with inferior survival outcomes in patients with NSCLC receiving immunotherapy, whereas probiotic supplementation shows a potential but yet unconfirmed benefit. These findings highlight the emerging role of gut microbiota as a host-associated determinant of immunotherapy efficacy and support the need for well-designed prospective studies to define microbiome-informed therapeutic strategies.

Author contributions

Conceived and designed the analysis: Dawei Chen, Qian Song, Jinming Yu.

Collected the data: Shuling Ma, Zijun Zhai, Shijie Shang, Xinyi Liang, Shan Yin.

Contributed data or analysis tools: Shuling Ma, Zijun Zhai, Rui Ding, Ran Zhang, Guomeng Sha.

Performed the analysis: Shuling Ma, Zijun Zhai, Haofeng Lin.

Wrote the paper: Shuling Ma, Zijun Zhai.