Jon posted a blog last week on mupirocin resistance in MRSA. This week, guest blogger Dr Gwen Knight (bio below) writes about a companion paper also published in the Journal of Antimicrobial Chemotherapy, which models mupirocin resistance…
It is a truth universally acknowledged, that acquiring most mechanisms of drug resistance incurs a fitness cost to the host bacterium. Determining the size of this cost and the impact that this cost will have on the spread of drug resistance is difficult. Is a 10% reduction in growth rate in the laboratory enough to stop resistance spreading in a hospital?
The answer is complex. Although resistant bacteria are likely to have a reduced growth rate, or fitness, in the absence of antibiotics, they have increased survival in the presence of antibiotics. If we continuously use high levels of antibiotic then resistance will be maintained. However, in the real world, there is an interplay between antibiotic usage and concentration, levels of resistance (i.e. low or high resistance) and fitness costs.
A recent modelling study explored this relationship by investigating the fitness costs associated with mupirocin resistance in MRSA. The authors used a detailed statistical analysis to estimate relative parameters from a clinical dataset and then inputted these into a simulation model to explore the impact of different control interventions for mupirocin resistant MRSA. From this they found that mupirocin resistance, in the absence of mupirocin exposure, has a large impact on the transmissibility of MRSA strains. This cost does not however, prevent resistance to mupirocin increasing under all of the explored intervention control measures explored such as ‘screen and treat’.
What was missing from this model analysis was much of the complexity in antibiotic resistance selection and transmission. For example, low and high mupirocin resistance was grouped together. We know from recent work that low level mupirocin resistance dynamics are different to those of high level mupirocin and it is likely, from experimental evidence, that they will have different fitness costs. Moreover, there was no inclusion of any other antibiotic resistances and the effect of differences in antibiotic exposure in different wards. I would suggest that MRSA should really stand for “many resistances S. aureus” and that only thinking about selection for mupirocin resistance in a hospital setting over simplifies what may be occurring. For example, processes such as bystander selection for MRSA or co-selection for other resistances on the plasmid encoding high level mupirocin resistance may be important.
From a broader perspective, we may also be thinking about ‘fitness’ inappropriately. We usually, as in this model, assume that resistant strains are less “transmissible”. Yet, as fitness is linked to the biological measure of growth rates, then potentially progression from colonisation to infection should also be delayed for drug resistant strains in our models.
Having said all of this, as a modeller myself, I can see that there are many strengths to this new study. For example there was an appropriate model structure and a detailed description of both it and the model parameters. Many were directly estimated from clinical data and a wide range of sensitivity analyses were performed. Most of the issues I have raised here could not be explored by the modellers due to insufficient data and the low number of MRSA isolates meaning that differentiation would be difficult. Importantly the paper concludes with irrefutable statements – they have provided evidence for a fitness cost to mupirocin resistance in the absence of antibiotic exposure and suggest that with mupirocin usage, mupirocin resistance will increase. Until modellers are provided with bigger datasets and epidemiological evidence relating to the differences in selection, this is probably the best we can do.
Bio: Dr Gwen Knight
Dr Gwen Knight is a mathematical modeller who is interested in modelling the spread of antibiotic resistance from both the bacterial and clinical viewpoint. She is currently a Career Development Fellow at Imperial College London within the NIHR funded HPRU on AMR and HCAI. Previously, she worked at the London School of Hygiene and Tropical Medicine on models to predict the impact of interventions to control the spread of M. tuberculosis. This interest continues into her current research where she is investigating fitness costs to resistance in M. tb in Peru, carbapenemase resistance in E. coli in the UK and developing models of bacterial population heterogeneity.