Scholarlink

<p>Infections remain a leading cause of death in acute leukaemia despite broad-spectrum empiric antibiotics. We hypothesized that antimicrobial resistance (AMR) follows predictable, linked patterns, making standard dual-coverage strategies often ineffective. We studied 115 febrile episodes in 80 adults at a North Indian tertiary centre. Phenotypic co-resistance was mapped using weighted networks, 3D Principal Component Analysis (PCA), and Louvain modularity. Sentinel antibiotics were identified using Eigenvector Centrality, and resistance patterns were correlated with clinical outcomes. Cultures (<i>n</i> = 81) showed high multidrug resistance: 42.5% carbapenem-resistant Gram-negative bacilli (CRO), 20% ESBL-producers, and 25.2% MRSA. ESBLs appeared earliest (median Day 0), MRSA on Day 2, and CRO later (Day 5.5; <i>p</i> = 0.027). Network analysis (on GNB isolates) revealed a Global Redundancy Score of 0.751, meaning ~ 75% of dual-coverage antibiotic combinations fail together. Ceftriaxone (centrality 0.463) and piperacillin-tazobactam (0.404) were sentinel hubs, predicting wider beta-lactam/inhibitor failure. PCA identified a multidrug-resistant core linking aminoglycosides and carbapenems (<i>r</i> = 0.60–1.00), with fluoroquinolones and doxycycline forming a separate salvage pathway. Hospital mortality was 17.2%, largely in AML (<i>p</i> &lt; 0.001). CRO infection was the strongest predictor of clinical failure (OR 6.33; 95% CI 1.52–26.34; <i>p</i> = 0.011). AMR in acute leukaemia follows a structured “architecture of failure.” Recognizing these co-resistance networks allows clinicians to front-load empiric coverage for early ESBL/MRSA, use carbapenems more judiciously, and reduce ineffective therapy, providing a practical, data-driven framework for high-risk patients.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Network Analysis of Antibiotic Co-Resistance in Gram-Negative Sepsis in Acute Leukaemia

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • Suvir Singh,
  • Eshani Dewan,
  • Veenu Gupta,
  • Monica Sahota,
  • Kunal Jain

摘要

Infections remain a leading cause of death in acute leukaemia despite broad-spectrum empiric antibiotics. We hypothesized that antimicrobial resistance (AMR) follows predictable, linked patterns, making standard dual-coverage strategies often ineffective. We studied 115 febrile episodes in 80 adults at a North Indian tertiary centre. Phenotypic co-resistance was mapped using weighted networks, 3D Principal Component Analysis (PCA), and Louvain modularity. Sentinel antibiotics were identified using Eigenvector Centrality, and resistance patterns were correlated with clinical outcomes. Cultures (n = 81) showed high multidrug resistance: 42.5% carbapenem-resistant Gram-negative bacilli (CRO), 20% ESBL-producers, and 25.2% MRSA. ESBLs appeared earliest (median Day 0), MRSA on Day 2, and CRO later (Day 5.5; p = 0.027). Network analysis (on GNB isolates) revealed a Global Redundancy Score of 0.751, meaning ~ 75% of dual-coverage antibiotic combinations fail together. Ceftriaxone (centrality 0.463) and piperacillin-tazobactam (0.404) were sentinel hubs, predicting wider beta-lactam/inhibitor failure. PCA identified a multidrug-resistant core linking aminoglycosides and carbapenems (r = 0.60–1.00), with fluoroquinolones and doxycycline forming a separate salvage pathway. Hospital mortality was 17.2%, largely in AML (p < 0.001). CRO infection was the strongest predictor of clinical failure (OR 6.33; 95% CI 1.52–26.34; p = 0.011). AMR in acute leukaemia follows a structured “architecture of failure.” Recognizing these co-resistance networks allows clinicians to front-load empiric coverage for early ESBL/MRSA, use carbapenems more judiciously, and reduce ineffective therapy, providing a practical, data-driven framework for high-risk patients.