Influence of microbial composition and sample type on antimicrobial resistance in urinary tract infections: a single-centre retrospective cohort study (2015-2023)

This single-centre retrospective study of nearly 190,000 urine cultures reveals that urine sample type and polymicrobial co-isolation significantly influence antimicrobial resistance patterns in urinary tract infections, suggesting that incorporating these contextual factors alongside pathogen identity could improve clinical reporting and empiric prescribing.

The Gist

The Big Picture: A Detective Story in the Bladder

Imagine the human bladder as a busy city square. Sometimes, this square gets invaded by unwanted guests (bacteria and fungi) causing an infection called a Urinary Tract Infection (UTI).

For decades, doctors have treated these infections like a "one-person crime." They usually look for the single most famous criminal (usually E. coli) and try to catch just that one.

This study is different. The researchers looked at nearly 190,000 urine samples from Zurich, Switzerland, over eight years. Instead of looking for a single criminal, they asked: "Who else is in the crowd? Does it matter if the crime happened in a quiet park (normal pee) or a high-security prison cell (a catheter tube)?"

They discovered that who is in the infection, where the sample came from, and who their friends are matters just as much as the main criminal itself.

---

1. The Three Types of "Crime Scenes" (Sample Types)

The study compared three ways of collecting urine, which act like different crime scenes:

• Midstream Urine (MU): This is a "clean" sample, like catching someone in a public park. It's usually just one or two intruders.
• Indwelling Catheter (IDC): This is a tube left in the bladder for a long time (like a permanent tent set up in the city square). This environment is a party for many different bacteria. The study found that infections here are 60% more likely to be "polymicrobial" (meaning a gang of different bacteria working together) compared to normal urine.
• Intermittent Catheter (IMC): This is a tube used occasionally (like a pop-up tent). It falls somewhere in between the park and the permanent tent.

The Analogy:
If a normal UTI is a mugging by a single thief, a catheter-associated UTI is a gang war. The "gang" includes bacteria that don't usually hang out together, like Pseudomonas (a tough, green-colored germ) and Candida (a type of yeast/fungus).

2. The "Gang Dynamics" (Co-occurrence)

The researchers used math to see which bacteria like to hang out together.

• Mostly Neutral: Most bacteria don't really care about each other; they just happen to be there because the environment (the catheter) is perfect for them.
• The "BFFs": Some pairs are best friends. For example, Candida albicans (yeast) and Candida glabrata (another yeast) almost always show up together in catheter samples.
• The "Enemies": Some bacteria avoid each other. E. coli and Pseudomonas rarely hang out together in the same sample.

The Analogy:
Think of the bacteria like people at a concert.

• In a normal urine sample, it's like a small coffee shop: mostly one type of person (E. coli), maybe a few others.
• In a catheter sample, it's like a massive music festival. You have different "tribes" (bacteria types) forming their own circles. Some tribes (like the Yeast gang) stick together tightly. Others avoid each other.

3. The "Super-Villain" Effect (Antibiotic Resistance)

This is the most critical part of the study. The researchers wanted to know: Does having a "gang" make the bacteria harder to kill with antibiotics?

• The Main Villain: The single biggest factor in whether a bacteria is resistant to drugs is what species it is. (e.g., E. coli is naturally tougher than others).
• The "Sidekick" Effect: However, the study found that if a "tough" bacteria (like E. coli) is hanging out with a "super-tough" partner (like Enterococcus faecium or Candida), the E. coli becomes even more resistant.

The Analogy:
Imagine a burglar (E. coli) trying to break into a house.

• Alone: He might be stopped by a standard lock (antibiotic).
• With a Partner: If he brings a friend who is a master of picking locks (Enterococcus), the burglar suddenly becomes much harder to catch. The presence of the partner seems to "boost" the main criminal's defenses.

4. The "Time Travel" Aspect (2015–2023)

The study also looked at how things changed over 8 years.

• The Trend: Resistance to antibiotics is going up. In 2015, about 48% of infections were resistant to drugs; by 2023, it was 60%.
• The Shift: Bacteria are getting better at resisting common drugs (like penicillin types) but are getting slightly worse at resisting older drugs (like fluoroquinolones).

The Analogy:
It's like an arms race. The bacteria are constantly upgrading their armor (resistance). The study shows that in the last few years, they have been upgrading their "penicillin shields" very quickly.

5. What This Means for You (The Takeaway)

The researchers conclude that doctors need to change how they think about UTIs.

1. Don't ignore the "Gang": If a urine test shows multiple bacteria, don't just treat the biggest one. The presence of the "partners" changes the risk.
2. Context Matters: A bacteria found in a catheter is a different "animal" than the same bacteria found in normal urine. Catheter bacteria are usually tougher and more resistant.
3. Better Prescriptions: By knowing who is in the infection and who they are hanging out with, doctors can choose the right antibiotic the first time, rather than guessing. This helps stop the spread of "superbugs."

Summary in One Sentence

This study proves that treating a urinary infection isn't just about identifying the main germ; it's about understanding the entire microbial neighborhood and how the catheter environment turns a simple infection into a complex, drug-resistant gang war.

Technical Summary

1. Problem Statement

Urinary tract infections (UTIs) represent a significant global healthcare burden, with catheter-associated UTIs (CAUTIs) accounting for approximately 40% of hospital-acquired infections. Current surveillance and clinical management face several limitations:

• Monomicrobial Bias: Most research and diagnostic protocols focus on single pathogens (primarily Escherichia coli), obscuring the dynamics of polymicrobial infections which are common in catheterized patients (up to 80% in long-term cases).
• Ecological Blind Spots: Standard surveillance often aggregates catheter and non-catheter urine samples, failing to distinguish device-specific ecological shifts from host factors.
• Resistance Drivers: The interplay between sample type (voided vs. catheterized), microbial community composition (co-isolation), and antimicrobial resistance (AMR) is poorly quantified in routine diagnostic data.
• Clinical Gap: There is a lack of data linking specific pathogen combinations and sample contexts to resistance profiles, hindering context-aware empiric prescribing.

2. Methodology

The study utilized a large-scale, retrospective cohort design analyzing 188,687 urine cultures from the Institute of Medical Microbiology at the University of Zurich (January 2015 – May 2023).

• Data Stratification: Samples were categorized into three types:
  • Midstream Urine (MU): 132,713 samples.
  • Indwelling Catheter (IDC): 31,975 samples.
  • Intermittent Catheter (IMC): 22,131 samples.
• Classification: Cultures were classified as negative, bacteriuria, or microbiologically significant UTI based on a threshold of ≥10⁵ CFU/mL.
• Statistical Framework:
  • Prevalence Analysis: Covariate-adjusted regression models (Prevalence Ratios) compared species abundance across sample types, adjusting for sex, age, ward, and year.
  • Co-occurrence Modeling:
    • Adjusted pairwise odds ratios (ORs) to identify non-random associations.
    • Degree-preserving permutation null models to distinguish true ecological interactions from artifacts of marginal prevalence.
    • Directional partner-choice models to analyze shifts in partner composition relative to MU samples.
  • Resistance Modeling: Antimicrobial resistance (AMR) was modeled as a Bernoulli outcome distinguishing between:
    • Total Resistance (TR): Acquired + Intrinsic resistance.
    • Acquired Resistance (AR): Resistance acquired via mutation or horizontal gene transfer (excluding intrinsic).
  • Feature Importance: Mutual Information was calculated to quantify the predictive power of organism identity versus patient metadata (sex, age, sample type) on resistance outcomes.

3. Key Contributions

• Ecological Framework: The study moves beyond single-pathogen analysis to treat UTIs as complex microbial communities, explicitly modeling how sample type (voided vs. catheter) reshapes the microbiome.
• Hierarchical Interaction Analysis: It employs a multi-layered statistical approach (PRs, ORs, permutations, and partner-choice models) to disentangle true biological co-occurrence from confounding factors like shared risk profiles.
• Quantification of Partner Effects: It provides empirical evidence that the presence of specific "partner" organisms (e.g., Enterococcus faecium or Candida albicans) significantly alters the resistance probability of a primary pathogen.
• Temporal Resistance Trends: It documents specific, non-seasonal shifts in resistance patterns over an 8-year period, highlighting rising $\beta$-lactam/$\beta$-lactamase inhibitor resistance.

4. Key Results

A. Microbial Composition and Polymicrobiality

• Polymicrobial Burden: Catheter-associated samples (IDC and IMC) were ~60% more likely to be polymicrobial than midstream urine (MU) samples (Adjusted PR 1.63).
• Pathogen Shifts:
  • E. coli was dominant in MU (39.8%) but significantly less prevalent in IDC (21.9%).
  • Device-associated enrichment: IDC samples showed strong enrichment for Pseudomonas aeruginosa, Candida albicans, and Candida glabrata.
  • Sex Differences: E. coli was more common in females, while Enterococcus faecalis and P. aeruginosa were enriched in males.
• Co-occurrence Patterns:
  • Most species pairs showed neutral or negative associations after adjustment.
  • Reproducible "Hotspots": Specific pairs showed consistent enrichment across methods, notably C. albicans–C. glabrata and C. albicans–E. faecium.
  • Asymmetry: The association between E. coli and E. faecalis was asymmetric; E. faecalis was more likely to co-occur with E. coli in IDC samples, but not vice versa, suggesting directional ecological benefits or shared risk factors.

B. Antimicrobial Resistance (AMR) Drivers

• Dominance of Organism Identity: Mutual information analysis revealed that organism identity was the dominant predictor of resistance (0.28 bits for TR, 0.15 bits for AR), far outweighing patient demographics or sample type (<0.01 bits each).
• Sample Type Effects:
  • IDC samples exhibited higher Total Resistance (TR) than MU.
  • Within common pathogens (e.g., E. coli, K. pneumoniae), Acquired Resistance (AR) was higher in IDC samples compared to MU.
• Partner Effects on Resistance:
  • Co-isolation with hospital-associated partners (E. faecium, C. albicans, P. aeruginosa) was associated with increased AR in primary pathogens across multiple drug classes (e.g., $\beta$-lactams, fluoroquinolones).
  • Conversely, co-isolation with low-virulence commensals (e.g., Actinotignum schaalii) was associated with reduced resistance in some contexts.
• Temporal Trends (2015–2023):
  • Overall AR increased from ~48% to ~60%.
  • Rising: Resistance to $\beta$-lactams and $\beta$-lactamase inhibitors (e.g., amoxicillin-clavulanate) in Enterobacterales.
  • Declining: Fluoroquinolone resistance (likely due to reduced usage).
  • Stable: Cephalosporin resistance (ESBL proxy) and carbapenem resistance remained low.

5. Significance and Clinical Implications

• Context-Aware Diagnostics: The study argues that sample type (catheter vs. voided) and community context (polymicrobial vs. monomicrobial) provide actionable information beyond pathogen identity. Ignoring these factors may lead to underestimating resistance risks in catheterized patients.
• Empiric Prescribing: The findings suggest that empiric therapy should be tailored not just to the isolated pathogen, but also to the sample source and the presence of co-isolated partners. For instance, a polymicrobial IDC sample containing E. coli and E. faecium may require broader coverage than a monomicrobial MU sample of E. coli.
• Stewardship: Understanding that specific ecological interactions (e.g., biofilm formation or cross-protection) may drive resistance highlights the need for stewardship strategies that target the microbial community rather than just single species.
• Future Directions: The authors call for prospective multicenter validation and the development of diagnostics that capture polymicrobial ecology (e.g., combined ASTs or culture-independent methods) to better address the growing AMR crisis in UTIs.

Limitations: The study is retrospective, single-center, and lacks symptom/treatment data, meaning some positive cultures may represent asymptomatic bacteriuria. Causality regarding resistance mechanisms cannot be established, only association.