Extensively drug-resistant gram-negative sepsis in a neonatal intensive care unit from western India: A retrospective cohort study

Methods

This retrospective study was undertaken by the department of Neonatology at All India Institute of Medical Sciences (AIIMS), Jodhpur, Rajasthan, India. The study was approved by the Institute’s ethics committee. Data were retrieved from the unit’s electronic database, and confidentiality was maintained.

Results

During the study period, 1230 neonates (excluding readmissions) were admitted to the NICU, 973 (79%) were inborn. Sepsis was diagnosed in 387 neonates (31.5%); 141 (36.4%) had culture-positive sepsis, representing 11.5% of total admissions. These neonates contributed 194 culture-positive sepsis episodes. GNB accounted for 143 (73.7%)isolates, of which 72.7% were MDR, including 33.6% identified as XDR (Figure).

Close

Figure.

Flow of study

Export to PPT

XDR Gram-negative sepsis occurred in 45 neonates (31.9% of culture-positive cases), contributing 48 episodes (24.7% of all isolates). The mean (±SD) gestational age and birth weight were 34±5 wk and 1831±803 g, respectively. Two-thirds were male, and 57.8% (n=26) were outborn referrals, with 60.7% (n=17) of isolates identified from blood cultures obtained within 48 h of admission (Table I). The median (IQR) postnatal age at culture positivity was 10.5 (5–23) days; 20.8% of episodes were early-onset. Compared with MDR and non-MDR groups, male sex and major congenital malformations were more frequent in the XDR group (Table I). Maternal clinical data were available for a subset of neonates with XDR Gram-negative bacterial sepsis. Histologically confirmed chorioamnionitis was identified in two mothers (Table II); in one case, XDR Klebsiella pneumoniae was isolated from both placental tissue and the neonatal blood culture.

All XDR sepsis episodes were clinically symptomatic, most commonly presenting with respiratory failure (n=30, 62.5%) and hemodynamic instability (n=24, 50%). Thrombocytopenia occurred in 56% (n=27), meningitis in 42% (n=20), and necrotising enterocolitis in 15% (n=7) of episodes. Escalation of respiratory support was required in 94% of episodes, inotropes in 48%, and blood product transfusion in 38%. In-hospital mortality among neonates with XDR GNB sepsis was 46.7% (Table II).

Overall, 87 neonates (7.1%) died; sepsis accounted for 67 deaths. Mortality was significantly higher in culture-positive compared with culture-negative sepsis (32.6% vs. 8.5%). Among gram-negative infections, mortality increased progressively from non-MDR (12.9%) to MDR (31.7%) and XDR sepsis (46.7%) (P=0.009) (Table III).

Klebsiella (43.7%) and Acinetobacter (37.5%) species predominated among XDR isolates, with substantially higher CFR compared to non-MDR strains (Table IV). All XDR isolates were susceptible to polymyxins, with limited susceptibility to other agents. A unit-specific antibiogram demonstrated persistently high XDR prevalence (20–29.7%) throughout the study period (Supplementary Tables II and III).

Discussion

This study demonstrates a high burden of antimicrobial resistance in neonatal sepsis, with Gram-negative bacterial sepsis accounting for nearly three-quarters of isolates, and one-third of these being XDR. Neonates with XDR Gram-negative sepsis had markedly higher mortality than those with MDR and non-MDR sepsis, underscoring the clinical impact of AMR in NICUs. Similar resistance patterns have been reported from multicentric surveillance studies in LMICs, including the Delhi Neonatal Infection Study (DeNIS) and the Young Infant Study (YIS), which documented AMR in 50–90% of neonatal Gram-negative isolates.^8,10 Likewise, the BARNARDS study, encompassing ∼36,000 infants across Africa and South Asia, reported that ∼90% of Gram-negative isolates were MDR or XDR.²³ In our cohort, ∼60% of XDR isolates were from outborn neonates, with three-fourths detected within 48 hours of admission, suggesting infection acquired prior to referral. Together with multicentric data from Indian district hospitals, these findings indicate that AMR is pervasive across neonatal care settings and not limited to tertiary centers.²⁴ In contrast, surveillance networks from high-income countries, such as ARPEC, report considerably lower MDR rates (10–30%), highlighting the disproportionate burden of AMR in neonatal units in LMICs.¹¹

Klebsiella and Acinetobacter species were the dominant XDR pathogens in this cohort, a pattern repeatedly observed in LMIC settings. Comparable studies from LMICs have reported XDR rates, ranging between 10-80% among Klebsiella isolates.^17,18 While the XDR is less frequently reported, MDR Gram-negative bacteria consistently remain the leading cause of neonatal sepsis.^8,11 The ARPEC study documented MDR in 58% of Klebsiella, 62.5% of Burkholderia, and 66.6% of Acinetobacter isolates.¹¹ Similarly, the DENIS study found MDR in 82% of Acinetobacter, 54% of Klebsiella, 50% of Enterobacter, and 38% of Escherichia isolates.⁸ Comparable resistance patterns, with MDR exceeding 70% in Klebsiella and Acinetobacter, have also been reported from South Asia.¹³ These organisms pose therapeutic challenges due to resistance to extended-spectrum cephalosporins and carbapenems, leaving few effective treatment options.

Despite timely diagnosis and appropriate supportive care, nearly half of the neonates with XDR Gram-negative sepsis in this study did not survive. While XDR sepsis was significantly associated with mortality, this relationship should not be interpreted as causal. Mortality in MDR and XDR sepsis is likely multifactorial, reflecting a complex interplay of pathogen virulence, baseline illness severity (particularly among outborn neonates), delayed initiation of effective antimicrobial therapy, and host vulnerability. In our unit, second-line empiric therapy includes vancomycin and meropenem; however, the high prevalence of carbapenem resistance among XDR isolates necessitated escalation to reserve agents such as colistin or polymyxins only after susceptibility results became available, typically after 48–72 h. This delay likely contributed to the higher mortality observed. Similar observations have been reported by ARPEC, which documented a fourfold higher mortality (34% vs. 8%) in MDR Gram-negative infections.¹¹ In contrast, DeNIS reported comparable mortality between MDR and non-MDR isolates (both ∼50%), possibly reflecting the high virulence of some susceptible pathogens in that cohort, though culture-positive sepsis with resistant isolates had five-fold higher odds of mortality compared to culture-negative sepsis.⁸

Currently, ample literature highlights the emergence of MDR strains and resistance to WHO-recommended first-line antibiotics.^5,25,26 Yet, data on XDR organisms in neonates remain limited, with most reports from adult ICUs showing prevalence rates between 3.5% and 15%.^17,27 The high endemicity of MDR and XDR gram-negative infections in LMICs is driven by gaps in antimicrobial and diagnostic stewardship, indiscriminate antibiotic use, reliance on invasive devices, and suboptimal infection control practices.^7,28-30 Additional contributors include indiscriminate antibiotic use in healthcare and agriculture, reinforcing the need for a coordinated ‘One Health’approach.^31,32 With increasing resistance to first-line and reserve antibiotics and stagnation in antibiotic development, preserving the effectiveness of existing agents through integrated stewardship strategies is imperative.^3,33

This study has several strengths. It is among the first to specifically focus on XDR GNB sepsis in neonates, an underreported yet critical public health concern. The use of automated blood culture systems, standardised resistance definitions, and follow-up until discharge enabled robust assessment of disease burden and outcomes. In addition, the development of a unit-specific cumulative antibiogram enhances the translational relevance of the findings for local empiric therapy and stewardship strategies.

This study has several limitations. The single-centre design in a tertiary NICU may limit generalisability; however, the predominance of outborn referrals partially mitigates this concern. The retrospective design resulted in incomplete data on prior antibiotic exposure, illness severity, and maternal or environmental factors, introducing potential ascertainment bias. Although mortality was higher in XDR sepsis, lack of data on time to appropriate therapy and the small number of outcome events precluded robust multivariable adjustment for key prognostic factors, limiting causal inference. Mortality, assessed as all-cause within 28 days, may overestimate sepsis-attributable deaths. The small number of XDR cases limited pathogen-specific analyses and the precision of estimates. Finally, the absence of data on healthcare-associated infections, molecular resistance mechanisms, and long-term neurodevelopmental outcomes constrains interpretation, highlighting the need for prospective, multicentric studies incorporating molecular diagnostics.

In conclusion, this study demonstrates a high and sustained burden of XDR Gram-negative bacterial sepsis in neonates, associated with increased mortality. These findings emphasise the need for strengthened surveillance, infection prevention strategies, and optimised antimicrobial and diagnostic stewardship to mitigate the impact of AMR in neonatal care.