A nonlinear time-series analysis approach to identify thresholds in associations between population antibiotic use and rates of resistance

Abstract

Balancing access to antibiotics with control of antibiotic resistance is a global public health priority. Currently, antibiotic stewardship is informed by a 'use it and lose it' principle, in which population antibiotic use is linearly related to resistance rates. However, theoretical and mathematical models suggest use-resistance relationships are non-linear. One explanation is that resistance genes are commonly associated with 'fitness costs', impairing pathogen replication or transmissibility. Therefore, resistant genes and pathogens may only gain a survival advantage where antibiotic selection pressures exceed critical thresholds. These thresholds may provide quantitative targets for stewardship: optimising control of resistance while avoiding over-restriction of antibiotics. We evaluated the generalisability of a nonlinear time-series analysis approach for identifying thresholds using historical prescribing and microbiological data from five populations in Europe. We identified minimum thresholds in temporal relationships between use of selected antibiotics and rates of carbapenem-resistant Acinetobacter baumannii (in Hungary), extended spectrum β-lactamase producing Escherichia coli (Spain), cefepime-resistant Escherichia coli (Spain), gentamicin-resistant Pseudomonas aeruginosa (France), and methicillin-resistant Staphylococcus aureus (Northern Ireland) in different epidemiological phases. Using routinely generated data, our approach can identify context-specific quantitative targets for rationalising population antibiotic use and controlling resistance. Prospective intervention studies restricting antibiotic consumption are needed to validate Results Identifying non-linear temporal relationships: from experiment to applicationIn a Monte Carlo experiment we compared the ability of linear and non-linear time-series analysis (Multivariate Adaptive Regression Splines, MARS) to identify pre-defined relationships between simulated explanatory and outcome time-series (Supplementary Figure 1). Non-linear time-series analysis (NL-TSA) accurately identified both truly linear and nonlinear associations. However, linear time-series analysis provided biased estimations and overall poorer data-fit if relationships were non-linear. NL-TSA models applied to retrospective time-series data from five European study populations (examples 1-5), frequently identified minimum thresholds in antibiotic useresistance relationships, (figures 1-5 and Supplementary Table 1). 'Ceiling effects', in which further increases in explanatory variables did not affect resistance rates, were found at highlevels of use of some antibiotics and hand hygiene. Non-linearities in autoregression and population interaction terms further indicated the complexity of transmission dynamics within and between clinical populations. Example 1: Carbapenem-resistant Acinetobacter baumannii (Debrecen, Hungary) We examined ecological determinants of carbapenem-resistant A. baumannii (CRAb) in a tertiary hospital population in Debrecen, Hungary (figure 1). Betwee...

Global Evolution of Pathogenic Bacteria With Extensive Use of Fluoroquinolone Agents

“…We believe the data presented in the paper demonstrate the crucial role fluoroquinolones played in the selection of the international clones/STs of MDR bacteria. Nevertheless, a variety of other groups of antibiotics are well-established to have also promoted the dissemination of MDR pathogens including strains of the international clones of HA-MRSA (Monnet et al, 2004;Lawes et al, 2015), MDR K. pneumoniae (da Silva et al, 2012;Ryu et al, 2018), MDR E. coli (López-Lozano et al, 2019) and C. difficile (Owens et al, 2008;Vardakas et al, 2016). However, the impact of these antibiotics remain second rate relative to that of fluoroquinolones as shown above.…”

Quantifying the economic cost of antibiotic resistance and the impact of related interventions: rapid methodological review, conceptual framework and recommendations for future studies

“…Advanced modelling approaches may circumvent some but not all of these limitations. Time-series approaches to modelling the relationship between population antibiotic use and ABR in hospitals suggest that this relationship is indeed non-linear [77]. Furthermore, machine learning approaches, such as boosted regression trees, can also be used, allowing flexibility in the functional relationship [70].…”

Mathematical basis for the assessment of antibiotic resistance and administrative counter-strategies

“…The time series analyses [8] supplied evidence that prevalence rates increase in a nonlinear fashion when exceeding a prevalence threshold after a sufficiently long duration of administration of certain antibiotics. The existence of such a threshold indicates that a switch to an alternative antibiotic agent is due.…”

“…[8]) that pointed to thresholds in associations between population antibiotic use and prevalence of resistant pathogens. Within the scope of the time series analyses as discussed by López-Lozano et al [8], the correlation of the time series of the prevalence of a specific pathogen with the time series of the amount of corresponding administered antibiotics is calculated. Commonly, such a correlation of two time series is given by a mutual entropy.…”