A hierarchical Bayesian latent class mixture model with censorship for detection of linear temporal changes in antibiotic resistance

Bayesian Calibration to Address the Challenge of Antimicrobial Resistance: A Review

Antimicrobial resistance (AMR) emerges when disease-causing microorganisms develop the ability to withstand the effects of antimicrobial therapy. This phenomenon is often fueled by the human-to-human transmission of pathogens and the overuse of antibiotics. Over the past 50 years, increased computational power has facilitated the application of Bayesian inference algorithms. In this comprehensive review, the basic theory of Markov Chain Monte Carlo (MCMC) and Sequential Monte Carlo (SMC) methods are explained. These inference algorithms are instrumental in calibrating complex statistical models to the vast amounts of AMR-related data. Popular statistical models include hierarchical and mixture models as well as discrete and stochastic epidemiological compartmental and agent based models. Studies encompassed multi-drug resistance, economic implications of vaccines, and modeling AMR in vitro as well as within specific populations. We describe how combining these topics in a coherent framework can result in an effective antimicrobial stewardship. We also outline recent advancements in the methodology of Bayesian inference algorithms and provide insights into their prospective applicability for modeling AMR in the future.

Salmonella enterica Serovar Dublin from Cattle in California from 1993 to 2019: Characterization and Analysis of Antimicrobial Resistance Diversity

For this study, antimicrobial susceptibility data for Salmonella enterica subsp. enterica serovar Dublin (S. Dublin)-a well-known cattle-adapted pathogen with current concerns for multidrug resistance-were recovered from cattle at the California Animal Health and Food Safety Laboratory System (CAHFS) over the last three decades (1993-2019) and were evaluated using tools to capture diversity in antimicrobial resistance. For this purpose, minimum inhibitory concentration (MIC) testing was conducted for 247 clinical S. Dublin isolates. Antimicrobial resistance (AMR) profiles revealed a predominant core multidrug-resistant pattern in the three most common AMR profiles observed. Antimicrobial resistance richness, diversity, and similarity analysis revealed patterns for changes in AMR profiles for different age groups. Discriminant analysis using MIC log2-transformed data revealed changes in MIC for year groups, with a time-sequence pattern observed. Drivers for reduced susceptibility were observed for 3rd generation cephalosporins and quinolones observed for more recent year groups (2011-2015 and 2016-2019) when compared to older year groups (1993-1999 and 2000-2005). Together, these results highlight the changes in the diversity of AMR profiles, as well as changes in susceptibility of S. Dublin over time for critical antimicrobials of importance to both animals and humans, and support the need for continued monitoring and efforts that will support judicious use of antimicrobials, especially for these two drug classes.

Effect of Group Housing of Preweaned Dairy Calves: Health and Fecal Commensal Antimicrobial Resistance Outcomes

The objectives of this study were to investigate the effects of group housing (three calves per group) on bovine respiratory disease (BRD), diarrhea and antimicrobial resistance (AMR) to fecal commensal Escherichia coli (EC) and enterococci/streptococci (ES). Our study comprised two arms, one experimental and one observational. In the experimental arm, preweaned calves on a California dairy were randomized to either individual (IND; n = 21) or group (GRP; n = 21) housing, using a modified California-style wooden hutch. The study period lasted from birth to 56 days of age, during which calves were health scored daily. Cumulative incidence and hazard ratios were estimated for disease. Antimicrobial resistance outcomes were assessed using a prospective cohort design; feces were collected from each calf three times per week and EC and ES were evaluated for AMR using the broth microdilution method against a panel of 19 antimicrobial drugs (AMD). Analysis of treatment records was used to select calves that had been exposed (EXP) to an AMD-treated calf. In GRP, exposure occurred when a calf was a hutchmate with an AMD-treated calf. In IND, exposure occurred when a calf was a neighbor with an AMD-treated calf (TRT). Age-matched unexposed calves (UNEXP) were then selected for comparison. Proportions of AMR in fecal commensals among EXP, UNEXP, and TRT calves were compared between GRP and IND. Accelerated failure time survival regression models were specified to compare differences in minimum inhibitory concentration (MIC) of fecal commensals between EXP and UNEXP calves within each of GRP and IND calves separately. Group calves had a BRD hazard 1.94 times greater that of IND calves (p = 0.03), using BRD treatment records as the outcome. For AMR in EC isolates, higher resistance to enrofloxacin was detected in enrofloxacin-EXP GRP isolates compared with enrofloxacin-EXP IND isolates, and UNEXP GRP calves had lower resistance to ceftiofur compared with enrofloxacin-EXP and enrofloxacin-TRT calves. A significant housing-by-time interaction was detected for EC ceftiofur MIC in EXP GRP calves at 4-14 days post exposure to enrofloxacin (MIC EXP-UNEXP: µg/mL (95% CI): 10.62 (1.17, 20.07)), compared with UNEXP calves. The findings of this study show an increase in BRD hazard for group-housed calves and an increase in ceftiofur resistance in group-housed calves exposed to an enrofloxacin-treated calf.

A multicategory logit model detecting temporal changes in antimicrobial resistance

Monitoring and investigating temporal trends in antimicrobial data is a high priority for human and animal health authorities. Timely detection of temporal changes in antimicrobial resistance (AMR) can rely not only on monitoring and analyzing the proportion of resistant isolates based on the use of a clinical or epidemiological cut-off value, but also on more subtle changes and trends in the full distribution of minimum inhibitory concentration (MIC) values. The nature of the MIC distribution is categorical and ordinal (discrete). In this contribution, we developed a particular family of multicategory logit models for estimating and modelling MIC distributions over time. It allows the detection of a multitude of temporal trends in the full discrete distribution, without any assumption on the underlying continuous distribution for the MIC values. The experimental ranges of the serial dilution experiments may vary across laboratories and over time. The proposed categorical model allows to estimate the MIC distribution over the maximal range of the observed experiments, and allows the observed ranges to vary across labs and over time. The use and performance of the model is illustrated with two datasets on AMR in Salmonella.

A Bayesian approach to modeling antimicrobial multidrug resistance

Multidrug resistance (MDR) has been a significant threat to public health and effective treatment of bacterial infections. Current identification of MDR is primarily based upon the large proportions of isolates resistant to multiple antibiotics simultaneously, and therefore is a belated evaluation. For bacteria with MDR, we expect to see strong correlations in both the quantitative minimum inhibitory concentration (MIC) and the binary susceptibility as classified by the pre-determined breakpoints. Being able to detect correlations from these two perspectives allows us to find multidrug resistant bacteria proactively. In this paper, we provide a Bayesian framework that estimates the resistance level jointly for antibiotics belonging to different classes with a Gaussian mixture model, where the correlation in the latent MIC can be inferred from the Gaussian parameters and the correlation in binary susceptibility can be inferred from the mixing weights. By augmenting the laboratory measurement with the latent MIC variable to account for the censored data, and by adopting the latent class variable to represent the MIC components, our model was shown to be accurate and robust compared with the current assessment of correlations. Applying the model to Salmonella heidelberg samples isolated from human participants in National Antimicrobial Resistance Monitoring System (NARMS) provides us with signs of joint resistance to Amoxicillin-clavulanic acid & Cephalothin and joint resistance to Ampicillin & Cephalothin. Large correlations estimated from our model could serve as a timely tool for early detection of MDR, and hence a signal for clinical intervention.

A Bayesian latent class mixture model with censoring for correlation analysis in antimicrobial resistance across populations

Background

The emergence of antimicrobial resistance across populations is a global threat to public health. Surveillance programs often monitor human and animal populations to evaluate trends of emergence in these populations. Many national level antibiotic resistance surveillance programs quantify the proportion of resistant bacteria as a means of monitoring emergence and control measures. The reason for monitoring these different populations are many, including interest in similar changes in resistance which might provide insight into emergence and control options.

Methods

In this research, we developed a method to quantify the correlation in antimicrobial resistance across populations, for the conventionally unnoticed mean shift of the susceptible bacteria. With the proposed Bayesian latent class mixture model with censoring and multivariate normal hierarchy, we address several challenges associated with analyzing the minimum inhibitory concentration data.

Results

Application of this approach to the surveillance data from National Antimicrobial Resistance Monitoring System led to a detection of positive correlation in the central tendency of azithromycin resistance of the susceptible populations from Salmonella serotype Typhimurium across food animal and human populations.

Conclusions

Our proposed approach has been shown to be accurate and superior to the commonly used naïve estimation by simulation studies. Further implementation of this Bayesian model could serve as a useful tool to indicate the co-existence of antimicrobial resistance, and potentially a need of clinical intervention.

Could simulation methods solve the curse of sparse data within clinical studies of antibiotic resistance?

Infectious disease (ID) physicians and ID pharmacists commonly confront therapeutic questions relating to antibiotic resistance. Randomized controlled trial data are few and meta-analytic-based approaches to develop the evidence-base from several small studies that might relate to an antibiotic resistance question are not simple. The overriding challenge is the sparsity of data which is problematic for traditional frequentist methods, being the paradigm underlying the derivation of ‘P value’ inferential statistics. In other sparse data contexts, simulation methods enable answers to key questions that are meaningful, quantitative and potentially relevant. How these simulation methods ‘work’ and how Bayesian-based methods, being not ‘P value based’, can facilitate simulation are reviewed. These methods are becoming increasingly accessible. This review highlights why sparse data is less of an issue within Bayesian versus frequentist paradigms. A fictional pharmacokinetic study with sparse data illustrates a simplistic application of Bayesian and simulation methods to antibiotic dosing. Whether within epidemiological projections or clinical studies, simulation methods are likely to play an increasing role in antimicrobial resistance research within both hospital and community studies of either rare infectious disease or infections within specific population groups.

Longitudinal Trajectories of Hair Cortisol: Hypothalamic-Pituitary-Adrenal Axis Dysfunction in Early Childhood

The objective of this study was to examine if longitudinal trajectories of hair cortisol concentrations (HCC) measured at two or three yearly time points can identify 1-3 year old children at risk for altered hypothalamic-pituitary-adrenal (HPA)-axis function due to early life stress (ELS). HCC was measured (N = 575) in 265 children using a validated enzyme-linked immunosorbent assay. Hair was sampled in Clinic Visits (CV) centered at years 1, 2, and 3 (n = 45); 1 and 2 (n = 98); 1 and 3 (n = 27); 2 and 3 (n = 95). Log-transformed HCC values were partitioned using latent class mixed models (LCMM) to minimize the Bayesian Information Criterion. Multivariable linear mixed effects models for ln-HCC as a function of fixed effects for age in months and random effects for participants (to account for repeated measures) were generated to identify the factors associated with class membership. Children in Class 1 (n = 69; 9% Black) evidenced declining ln-HCC across early childhood, whereas Class 2 members (n = 196; 43% Black) showed mixed trajectories. LCMM with only Class 2 members revealed Class 2A (n = 17, 82% Black) with sustained high ln-HCC and Class 2B (n = 179, 40% Blacks) with mixed ln-HCC profiles. Another LCMM limited to only Class 2B members revealed Class 2B1 (n = 65, 57% Black) with declining ln-HCC values (at higher ranges than Class 1), and Class 2B2 (n = 113, 30% Black) with sustained high ln-HCC values. Class 1 may represent hair cortisol trajectories associated with adaptive HPA-axis profiles, whereas 2A, 2B1, and 2B2 may represent allostatic load with dysregulated profiles of HPA-axis function in response to varying exposures to ELS. Sequential longitudinal hair cortisol measurements revealed the allostatic load associated with ELS and the potential for developing maladaptive or dysregulated HPA-axis function in early childhood.

Overview of Quantitative Methodologies to Understand Antimicrobial Resistance via Minimum Inhibitory Concentration

Simple Summary

An emerging threat to human and food animal health is the development of antimicrobial resistance in bacteria associated with food animals. One of the primary tools for assessing resistance levels and monitoring for changes in expressed resistance is the use of minimum inhibitory concentration tests, which expose bacterial isolates to a series of dilutions of an antimicrobial agent to identify the lowest concentration of the antimicrobial that effectively prevents bacterial growth. These tests produce a minimum inhibitory value that falls within a range of concentrations instead of an exact value, a process known as censoring. Analysis of censored data is complex and careful consideration of methods of analysis is necessary. The use of regression methods such as logistic regression that divide the data into two or three categories is relatively easy to implement but may not detect important changes in the distributions of data that occur within the categories. Models that do not simplify the data may be more complex but may detect potentially relevant changes missed when the data is categorized. As a result, the analysis of minimum inhibitory concentration data requires careful consideration to identify the appropriate model for the purpose of the study.

Abstract

The development of antimicrobial resistance (AMR) represents a significant threat to humans and food animals. The use of antimicrobials in human and veterinary medicine may select for resistant bacteria, resulting in increased levels of AMR in these populations. As the threat presented by AMR increases, it becomes critically important to find methods for effectively interpreting minimum inhibitory concentration (MIC) tests. Currently, a wide array of techniques for analyzing these data can be found in the literature, but few guidelines for choosing among them exist. Here, we examine several quantitative techniques for analyzing the results of MIC tests and discuss and summarize various ways to model MIC data. The goal of this review is to propose important considerations for appropriate model selection given the purpose and context of the study. Approaches reviewed include mixture models, logistic regression, cumulative logistic regression, and accelerated failure time–frailty models. Important considerations in model selection include the objective of the study (e.g., modeling MIC creep vs. clinical resistance), degree of censoring in the data (e.g., heavily left/right censored vs. primarily interval censored), and consistency of testing parameters (e.g., same range of concentrations tested for a given antibiotic).