Global forecast of antimicrobial resistance in invasive isolates of Escherichia coli and Klebsiella pneumoniae

Highlights • We aimed to estimate global trends of antimicrobial resistance of E. coli and K. pneumoniae invasive isolates.• By 2030, over half of E. coli and K. pneumoniae invasive isolates could become resistant to third generation cephalosporins.• Carbapenem resistance is increasing rapidly among K. pneumoniae invasive isolates.


Introduction
Antimicrobial resistance (AMR) is now a global problem, and resistance in Enterobacteriaceae, specifically Escherichia coli and Klebsiella pneumoniae, is a critical threat to human health (World Health Organization, 2014;WHO, 2018). Infections caused by third-generation cephalosporin-resistant (3GCR) Enterobacteriaceae are associated with increased mortality, length of stay, and costs compared with drug-sensitive strains (Stewardson et al., 2016). Carbapenems are less reliable as last-resort antibiotics because of increasing resistance (Gelband et al., 2015). AMR already imposes a heavy economic burden on health systems (Stewardson et al., 2016). Projecting future prevalence of AMR may help prioritize research projects and interventions. The aims of this study were to estimate global trends in AMR in E. coli and K. pneumoniae and to project future AMR prevalence to 2030.

Methods
Data on population and gross national income per capita (GNIPC) from the World Bank and on AMR from ResistanceMap, a global repository of AMR data from quality-assured and accredited hospitals and laboratory networks, were used (CDDEP, 2018). Countries for which samples were obtained from a single hospital were excluded from this study. Annual AMR data that had fewer than 30 isolates were also excluded (Agresti and Caffo, 2000).
With few exceptions, low-and middle-income countries are less likely to monitor AMR; therefore, high-income countries are overrepresented in the ResistanceMap database. Not taking this into account may lead to selection bias and an underestimation of the prevalence of AMR because of the strong negative association between GNIPC and the prevalence of AMR (Alvarez-Uria et al., 2016). To overcome this problem, inverse probability of inclusion (IPI) weighting, a method analogous to the use of inverse probability weights, was used to account for non-responders in surveys (Dugoff et al., 2014). IPI weights were calculated based on the inverse probability of being included in the study, using a logistic regression model that included data from countries in the world for which GNIPC data were available (countries with no GNIPC data comprised 1.3% of the world population) (The World Bank, 2018). In this logistic regression model, the availability of national AMR data (thus being included in the study) was the dependent variable, and orthogonal cubic spline transformations of 2014 GNIPC and 2014 country populations were the independent covariates (Dugoff et al., 2014). IPI weights gave more 'weight' to countries that were less likely to have AMR data in the ResistanceMap database, based on their GNIPC and population. This method helps generalize the results of the study to the world population. IPI weights were multiplied by population weights, which gives more weight to countries with larger populations, and the results were used as probability or sample weights in the final mixed model with random intercept and slopes (Dugoff et al., 2014). The mixed models were used to project AMR up to 2030.

Results
The study included 45 countries with AMR data for E. coli and 43 countries with AMR data for K. pneumoniae. In countries with E. coli AMR data, the median number of AMR point estimates was 14 (interquartile range 1-15), and 31 were high-income countries. In countries with K. pneumoniae AMR data, the median number of AMR point estimates was 10 (interquartile range 2-14), and 28 were high-income countries. No country had AMR data beyond 2015.

Discussion
The projections of AMR in this study signal a potentially serious shortage of effective antimicrobials for common causes of infection by 2030. Under current trends, over three-fourths of E. coli globally will be 3GCR, and over half of K. pneumoniae invasive isolates will be CR. The consequences of the high prevalence of AMR could be devastating for health systems (World Health Organization, 2014;de Kraker et al., 2011).
The models showed that the annual variation in the prevalence of 3GCR K. pneumoniae was not significantly different from zero, with narrowing of the confidence interval over time. This can be explained by the fact that countries with initial low prevalence of 3GCR K. pneumoniae showed a rising trend over time, while the trends were stable or mildly decreasing in countries with initial high prevalence, such as India and South Africa (Supplementary material, Figure S3). CR K. pneumoniae had the highest annual increase of AMR, which could reach 53% by 2030, but the confidence intervals were wide, indicating uncertainty of the projections. The projected increase in the prevalence of CR E. coli was more modest. However, empirical treatment of infections will need to cover 3GCR, leading to an increased use of carbapenems, and this, in turn, may accelerate the pace of CR.
Enterobacteriaceae are part of the human gut microbiota, and the spread of AMR is facilitated by conditions that are more common in resource-poor settings, such as suboptimal sewage systems and a lack of access to clean water (Holmes et al., 2016). Previous studies have shown that resistance in Enterobacteriaceae can emerge anywhere and spread around the globe (Nordmann et al., 2011). Isolated interventions in high-income countries alone, without intervention efforts in low-and middle-income countries, may be ineffective in a globalized world (Nordmann et al., 2011).
This study has important limitations. The total population of all countries included in the study was approximately a third of the world population and was biased towards high-income countries. While IPI models were used to attempt to correct for underrepresentation of low-and middle-income countries, more surveillance data are urgently needed to improve current and future estimates of AMR. The projections for future levels of AMR were based on linear models, which assumed no changes in the growth rate of resistance. They also did not account for saturation or stabilization of AMR levels, as was observed with 3GCR K. pneumoniae. In addition, it was not possible to distinguish the case mix of community-and hospital-acquired infections among the countries included in the study, and the high prevalence of AMR in some countries could be influenced by a higher proportion of hospitalacquired infections (Dat et al., 2017;Thaden et al., 2017).
These results suggest that if current trends were to continue, third-generation cephalosporins and carbapenems could become ineffective against E. coli and K. pneumoniae in most parts of the world in the not-too-distant future. Empirical antimicrobial therapy for sepsis or for urinary tract or abdominal infections might shift to non-beta-lactam antibiotics, which, in turn, may lead to an increase in AMR in other antibiotic groups. These results underscore the need to improve the judicious use of antimicrobials and support recent World Health Organization recommendations to prioritize the research, discovery, and development of new and effective antibiotic treatments for beta-lactam-resistant Enterobacteriaceae (WHO, 2018).

Funding
The funders had no role in the study design, in the collection, analysis, and interpretation of the data, in the writing of the report, or in the decision to submit the article for publication.

Ethics approval
This study used data available in the public domain and thus did not require ethics approval.

Conflict of interest
There are no conflicts of interest to disclose.