Pharmacoepidemiological approaches to examining antimicrobial related adverse events in older adults

: (Alternative Format Thesis)

Abstract

Introduction

Older adults (aged 65 years and above) constitute the fastest growing population cohort in the western world. There is increasing evidence that the burden of infections disproportionately affects older adults, and hence this vulnerable population is frequently exposed to antimicrobials. These antimicrobials induce a differential risk of adverse drug events, including acute kidney injury (AKI), in this cohort. Detecting and predicting the adverse antimicrobial events, in this population cohort, is faced with several challenges including frequent, and sometimes inappropriate polypharmacy.



Aim

The aim of this PhD was to characterise antimicrobial-associated adverse events in older adults from a population perspective, employing pharmacoepidemiologic approaches to understand the extent of these risks. The objectives were to conduct a systematic review and meta-analysis for the risk of antibiotic-associated organ injury in older adults (65 years or above) following exposure to antibiotics; uncover antibiotics associated with acute kidney injury among older adults at a population level, using disproportionality analysis methods on post-marketing surveillance data; and to employ the case-crossover design to detect antibiotics associated with acute kidney injury among older adults on population datasets from New Zealand and the United Kingdom.



Methods

This thesis comprises of four empirical elements:

A systematic review and meta-analysis.

Application of disproportionality analysis methods on post-marketing surveillance data.

Application of the case-crossover design to detect antibiotics associated with acute kidney injury using a New Zealand dataset.

Application of the case-crossover design to detect antibiotics associated with acute kidney injury using a United Kingdom dataset.

For the systematic review and meta-analysis, we searched for original research articles in PubMed, Embase.com, Web of Science core collection, Web of Science BIOSIS citation index, Scopus, Cochrane Central Register of Controlled Trials, ProQuest, and PsycINFO databases. Key words in titles and abstracts, in addition to using MeSH terms, were used for search strings. We searched for all available articles up to 31st May 2021. After removing duplicates, articles were screened using inclusion and exclusion criteria from the study by two reviewers. The Newcastle-Ottawa scale was used to assess the risk of bias for cohort and case-control studies. We explored the heterogeneity of the included studies using the Q test and I2 test and the publication bias using the funnel plot and Egger's test. The meta-analyses were performed using the OpenMetaAnalyst software.

A case/non-case method was used to assess AKI risk associated with antibacterials between 1st January 2000 and 30th September 2021. Cases were Individual Case Safety Reports (ICSRs) for antibacterials, with AKI as preferred terms, included in the Medical Dictionary of Regulatory Activities (MedDRA) system organ classes, 'Renal and urinary disorders' disorders. The analyses were completed on a de-duplicated dataset containing only the recent version of the ICSR.  Signals were defined by a lower 95% confidence interval (CI) of reporting odds-ratio (ROR) ≥ 2, proportional reporting ratio (PRR) ≥ 2, information component (IC) > 0, Empirical Bayes Geometric Mean (EBGM) > 1, and reports ≥ 4. Sensitivity analyses were conducted a priori to assess the robustness of signals.  

For the case-crossover design using the New Zealand dataset, the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification code N17 was used to identify all individuals aged 65 years and above with a diagnosis of incident AKI on admission between 1st January 2005, and 31st December 2020, from the New Zealand National Minimum Data Set. A case-crossover cohort for antibiotic exposures, with a 3-day case period and two 30-day washout periods, summed up to a 66-day study period, was created. Using conditional logistic regression, the changed odds of AKI due to exposure to an antibiotic was calculated as matched odds ratios and their 95% confidence intervals. 

The UK Biobank was created by multiple stakeholders in response to the challenge of understanding the determinants of common complex diseases. More than 500,000 adults aged from 38 years to 70 years were recruited between 2006 and 2010, and extensive baseline assessments were carried out. The ICD-10 codes N17.0, N17.1, N17.2, N17.8, N17.9, and Read Code K040 were used to identify all individuals aged 65 years and above with a diagnosis of incident AKI on admission from the UK Biobank dataset. A case-crossover cohort for antibiotic exposures, with a 3-day case period and two 30-day washout periods, summed up to a 66-day study period, was created. Using conditional logistic regression, the changed odds of AKI due to exposure to an antibiotic was calculated as matched odds ratios and their 95% confidence intervals.



Results



Systematic review and meta-analysis

From the systematic review and meta-analysis, the overall absolute risks of acute kidney injury among older adults prescribed aminoglycosides, glycopeptides, and macrolides were 15.1% (95% CI: 12.8-17.3), 19.1% (95% CI: 15.4-22.7), and 0.3% (95% CI: 0.3-0.3), respectively. Only 3 studies reported antimicrobial associated drug-induced liver injury. Studies reporting on the association of organ injury and antimicrobial exposure by age or duration of treatment were too few to perform a meta-analysis. The funnel plot and Egger's tests did not indicate evidence of publication bias.



Disproportionality analysis

A total of 3,680,621 reports on adverse drug events (ADEs) were retrieved from FDA Adverse Event Reporting System (FAERS) over the study period, of which 92,194 were antibacterial reports. Gentamicin, sulfamethoxazole, trimethoprim, and vancomycin consistently gave strong signals of disproportionality on all four disproportionality measures and across the different sensitivity analyses: gentamicin (ROR = 2.95[2.51-3.46]), sulfamethoxazole (ROR = 2.97[2.68-3.29]), trimethoprim (ROR = 2.81[2.29-3.46]), and vancomycin (ROR = 3.35[3.08-3.64]). 

Case-crossover study: New Zealand dataset.

A total of 2399 incident cases of AKI were identified between 2005 and 2020 among older adults. The adjusted odds of consuming sulfamethoxazole/trimethoprim antibiotic during the case period were 3.57 times (95% CI: 2.86 to 4.46) higher than the reference period among the incident AKI cases. Fluoroquinolone utilisation was also associated with incident AKI (adjusted OR = 2.56; 95% CI: 1.90 to 3.46). 



Case-crossover study: United Kingdom dataset

Among the 2006 participants with acute kidney injury, only 265 had an admission diagnosis of AKI. The adjusted odds for consuming amphenicols during the case period was 1.5 times (95% CI: 0.53 to 4.21) higher than the reference period among the incident AKI cases, while the matched odds ratio for imidazole was 0.33 (95% CI: 0.04 to 3.21). Due to the very low numbers of cases that had an exposure to antibiotics during the case period, our results did not detect any association between antibiotic classes and AKI.



Conclusion

This thesis highlights the complexity of antibiotic-associated AKI risks in older adults, demonstrating how these risks may vary based on health status, antibiotic class, and the population under study. The findings underscore the importance of vigilant prescribing practices, particularly in frail populations, while also recognising the critical need to balance antibiotic safety with the necessity of treating life-threatening infections.

Date of Award	8 Oct 2025
Original language	English
Awarding Institution	University of Bath
Supervisor	Prasad Nishtala (Supervisor) & Sandipan Roy (Supervisor)