Relationship between antibotic use and resistance: Shall we talk about thresholds and quotas?

By Dr. Jose María López-Lozano MD, PhD

In a worrisome scenario where not enough new antibiotics are expected to be available in the near future, it is critical to investigate approaches that help us to optimize the use of currently available antibiotics. In this post I will present a scientific proposal trying to explain the relationship between antibiotic use and resistance from a different perspective that might help us to better understand this phenomenon. This proposal is based on the findings of 3 research articles that have been published in the last couple of years (1-3). Assuming that this proposal needs to be contrasted and that its reproducibility needs to be verified I would be very happy if it could be discussed here, the ESGAP blog.

 

Time Series Analysis (TSA) techniques were used to study the relationship between antibiotic use and resistance as early as in 2000. The rationale was that resistance, measured over time and from an ecological point of view, is a stochastic phenomenon that results from the dynamic interaction of several factors, which, in turn, are also stochastic (use of antibiotics, spontaneous modifications of bacterial flora, hygiene and infection control measures , etc.) (4). That first proposal was based on a linear conception of the relationship between the triggers factors and their outcome, resistance: that is to say, the more antibiotic use, the more resistance, regardless of the level or intensity of use.

Stuart Levy(5), in 1994, hypothesized that such a relationship might not be linear. He suggested that there might be a threshold of antibiotic use beyond which, resistance would be triggered. On the other hand, below that given threshold or level of antibiotic use resistance would remain at infraepidemic levels, as a sporadic phenomenon. As far as we know, so far no one had tried to explore the ​​Levy’s hypothesis, nor to model or quantify it.

 

In the three papers above mentioned, we introduced a statistical methodology, from the field of Econometrics, suitable for the identification and estimation of nonlinear models. This is what is known as Multivariate Adaptive Regression Splines (MARS), based on the separation of the data into sections or “regions” in which the ratio of the explanatory variables to the dependent variable changes and allows the identification of the nodes in that change occurs. This statistical approach has allowed us to detect multiple situations in which, up to a certain threshold, no relationship is detected between the use of antibiotics but, beyond that threshold, the relationship is positive.

 

Likewise, if we were able to detect thresholds for all antibiotics used in a particular hospital, we could establish a policy of use aimed at not exceeding those thresholds, in the hope that resistance levels would remain at acceptable levels. For example, establishing quotas (max number of treatable patients) in order to remain under the threshold(3):

 

Dr. Jose María López-Lozano MD, PhD

Infection Control Team

Hospital Vega Baja. Orihuela (Spain)

 

 

References:

  1. Lawes T, Lopez-Lozano JM, Nebot CA, Macartney G, Subbarao-Sharma R, Wares KD, Sinclair C, Gould IM. Effect of a national 4C antibiotic stewardship intervention on the clinical and molecular epidemiology of Clostridium difficile infections in a region of Scotland: a non-linear time-series analysis. The Lancet Infectious Diseases , Volume 17 , Issue 2 , 194 – 206
  1. Lawes T, Lopez-Lozano JM, Nebot CA, Macartney G, Subbarao-Sharma R, Dare CR, Wares KD Gould IM. Effects of national antibiotic stewardship and infection control strategies on hospital-associated and community-associated meticillin-resistant Staphylococcus aureus infections across a region of Scotland: a non-linear time-series study. The Lancet Infectious Diseases , Volume 15 , Issue 12 , 1438 – 1449
  1. Lawes T, López-Lozano J-M, Nebot C, et al. Turning the tide or riding the waves? Impacts of antibiotic stewardship and infection control on MRSA strain dynamics in a Scottish region over 16 years: non-linear time series analysis. BMJ Open. 2015;5(3):e006596. doi:10.1136/bmjopen-2014-006596.

 

  1. López-Lozano JM, Monnet DL, Yagüe A et al. Modelling and forecasting antimicrobial resistance and its dynamic relationship to antimicrobial use: a time series analysis. Int J Antimicrob Agents 2000;14:21–31

 

  1. Levy SB. Balancing the drug-resistance equation. Trends Microbiol 1994;2:341–2

3 Comments

  • Conan MacDougall says:

    I think this is a brilliant approach and it’s so heartening seeing such sophisticated analyses applied to what is such a complex and almost surely non-linear problem. I wonder whether there would be both a “threshold effect” and a “time effect” – early in the lifetime of the antibiotic, the threshold for effect might be higher but as the years advance and low-level resistance accumulates in the “environment”, that threshold might lower for a particular time period (e.g. yearly).

    I also wonder about the potential for applying non-linear analysis to the study of the impact of exposure on risk on an individual patient-level. For example, what is the additional marginal effect of further exposure (duration of treatment) on the risk of emergence of resistance among GI flora?

    • (Dr) Timothy Lawes says:

      Dear Dr MacDougall,

      many thanks for your positive comments on the applications of our methodology. I would reiterate Dr López-Lozano’s general point that in this nascent field of understanding the population-level dynamics of antimicrobial use and resistance there is much still left to learn.

      If i understand your point about a ‘time effect’ correctly, you suggest that the threshold may change over time depending on the specific molecular epidemiology of the period. There are sound reasons for considering this hypothesis in addition to the point you make about accumulating low-level resistance. According to the ‘total use threshold’ theory, the initial threshold will occur where antibiotic selection pressure is high enough that the survival advantages of an acquired resistance outweigh its associated ‘fitness cost’. It is known that ‘fitness costs’ of resistance genes/mutations are highly variable so the initial level may depend upon which resistance genes are prevalent at that time. However, under constant high antibiotic selection pressure genes/mutations that confers equal antibiotic resistance at lower ‘fitness cost’ could be favoured. By way of example, the dominant MRSA strain in UK hospitals shifted from CC30 (EMRSA-16) to the ‘fitter’ CC22 (EMRSA-15) strain when CC22 acquired multiple new resistances. In addition secondary compensatory mutations may reduce the fitness cost of the original mutation. This means that over time the tipping point between antibiotic selection pressure and fitness costs change and the threshold for controlling resistance falls. The implication is that we may have to suppress population antibiotic use to levels below the original threshold at which resistance emerged in order to achieve control.

      It should be emphasised that one of the tenants of our methodology is that robust decisions about antimicrobial stewardship requires understanding epidemiology in specific populations at specific times. A benefit of the non-linear time-series analysis procedure is that it is relatively efficient. It is based upon data that is increasingly available in electronic formats from laboratories, pharmacies, and surveillance units. This means it can be iteratively applied to inform decisions in changing circumstances just as it does in the complex dynamic systems of economics and environmental science.

  • ESGAP-OVLC admin says:

    Thank you, Dr. MacDougall.

    Please see, Dr. JM López-Lozano reply here (http://esgap.escmid.org/?p=1505)

Leave a Reply