Understanding the factors that influence the distribution of - - PowerPoint PPT Presentation

Understanding the factors that influence the distribution of antibiotic resistance.

PHDL Seminar November 12th 2018 Derek MacFadden MD FRCPC
 *No Relevant Disclosures/Conflicts of Interest

Objectives

• 1. Review factors related to distribution of antibiotic
• resistance. 

• 2. Review the evidence supporting the role of climate
• n the population level distribution of antibiotic

resistance.


• 3. Review estimates of burden of antibiotic resistance

and how this is (and might be) measured.

Marston et al. Antimicrobial Resistance. JAMA. 316(11):1193-1204.

Emergence of Antibiotic Resistance

Projected Burden of Resistance

O’Neill. UK-AMR Review. https://amr-review.org/

Antibiotic Resistance is OneHealth

http://www.cdc.gov/drugresistance/threat-report-2013

Antibiotic Resistance is Heterogeneous by Region

CPE Endemicity 2010-2015

Albiger et al. Eurosurveillance 2015.

Antibiotic Resistance is Driven By Antimicrobial Use (Selection)

Hicks et al. NEJM 2013.

2010 Outpatient All ages All drugs #RX per 1000

Epidemiology of Antibiotic Resistance

• 1. Surveillance - generate a database of

geographically distributed measures of antibiotic resistance 


• 2. Evaluate factors associated with the

population distribution of antibiotic resistance

www.resistanceopen.com

Epidemiology of Antibiotic Resistance

• 1. Surveillance - generate a database of

geographically distributed measures of antibiotic resistance 


• 2. Evaluate factors associated with the

population distribution of antibiotic resistance

www.resistanceopen.com

http://www.resistanceopen.org

http://www.resistanceopen.org

Digital Surveillance

MacFadden et al. J Infect Dis 2016.

Surveillance Metrics

Current Status:


• 50 Countries
• >1700 Indices
• >10.6 million Isolates
• 2012-2017

In Progress:


• Metagenomic Data
• High Resolution States/

Countries

• ProMED Collaboration

Epidemiology of Antibiotic Resistance

• 1. Surveillance - generate a database of

geographically distributed measures of antibiotic resistance 


• 2. Evaluate factors associated with the population

distribution of antibiotic resistance

Epidemiology of Antibiotic Resistance

• 1. Surveillance - generate a database of

geographically distributed measures of antibiotic resistance 


• 2. Evaluate factors associated with the

population distribution of antibiotic resistance

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• 1. Antibiotic Use
• 2. Geography?
• 3. Population Factors?
• 4. Care setting?
• 5. Environmental/Climate Factors?

Goossens et al. Outpatient antibiotic use in Europe. Lancet. 2005.

We have a poor understanding of population level determinants of antibiotic resistance.

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• 1. Antibiotic Use
• 2. Geography?
• 3. Population Factors?
• 4. Care setting?
• 5. Environmental/Climate Factors?

Goossens et al. Outpatient antibiotic use in Europe. Lancet. 2005.

We have a poor understanding of population level determinants of antibiotic resistance.

What About Climate?

• 1. Temperature is one of the major drivers of


bacterial growth.

• 2. Seasonality in infection with Gram-negative


and Gram-positive infections = Carriage?

• 3. Horizontal gene transfer is typically


temperature dependent.

• 4. Suggestion of latitude gradients, 


typically attributed to antibiotic use.

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• Evaluated prevalence of resistance to routinely

tested antibiotics in E. coli, K. pneumoniae, S. aureus across the continental United States.


• ResistanceOpen, US Census, NOAA Climate,

CDC prescribing.

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• Evaluated prevalence of resistance to routinely

tested antibiotics in E. coli, K. pneumoniae, S. aureus across the continental United States.


• ResistanceOpen, US Census, NOAA Climate,

CDC prescribing.

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• Historical Minimum Temperature - parallels use in 


vector-borne diseases - ecologic suitability

• Population Density
• Antibiotic Prescribing Rates
• Reported Laboratory Standard
• Outpatient/Inpatient Sources

Predictors

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• 1.6 million human bacterial pathogens

• 41 States, 223 facilities

• 2013-2015

Dataset

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• Population level comparison

• Univariate associations with resistance prevalence

• Multivariable weighted regression models

Analysis

Antibiotic Resistance in E. coli

Amoxicillin

MacFadden et al. Antibiotic Resistance Increases with Local Temperature. Nature Climate Change. 2018.

Antibiotic Resistance in E. coli

All tested antibiotics

Resistance

• Min. Temperature

Antibiotic Resistance in K. pneumoniae

Septra

Antibiotic Resistance in K. pneumoniae

All tested antibiotics

Resistance

• Min. Temperature

Antibiotic Resistance in S. aureus

Cloxacillin

Antibiotic Resistance in S. aureus

All tested antibiotics

Resistance

• Min. Temperature

Significant Predictors of Resistance

Significant Predictors of Resistance

Cloxacillin: Adjusted Min Temp effect estimate -> 0.58 (p<0.0001)

Change Over Time

Summary

• Antibiotic prescribing, population density, and

temperature are associated with increased antibiotic resistance for common pathogens.


• The relationship between temperature and

antibiotic resistance may be increasing over time?

Limitations

• Capturing relevant time periods/measures for


antibiotic prescribing


• Population level data

• Confounding

Follow Up

• How do we validate these findings?

• Different region

• Longitudinal data

• Best possible AMR and AM consumption data

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• Evaluated prevalence of resistance to routinely

tested antibiotics in E. coli, K. pneumoniae, S. aureus across Europe.


• EARS-NET (ECDC), ESAC-NET (ECDC),

MERRA-2.

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• 4.5 million human bacterial pathogens

• 28 Countries across Europe

• Spanning 2000-2016

Dataset

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• Temperature Association (previous study) 

• Longitudinal Model (country specific intercepts)

• Evaluate rates of change as explanation for

geographic distribution Analysis

What Factors Impact on Distribution of 
 Antibiotic Resistance?

• Annual Minimum Temperature
• Population Density
• Antibiotic Consumption Rates
• Country Specific Intercepts
• Time

Predictors

McGough et al. BioRxiv 2018.

Escherichia coli

Escherichia coli

Klebsiella pneumoniae

Klebsiella pneumoniae

Staphylococcus aureus

Staphylococcus aureus

Summary

• Antibiotic resistance is generally increasing
• ver time in European countries.

• Antibiotic prescribing and population density

are associated with antibiotic resistance.


• Antibiotic resistance is increasing more rapidly

in warmer countries.

WHY?

http://www.cdc.gov/drugresistance/threat-report-2013

WHY?

• Season/temperature associated with carriage of

resistant Gram-negatives.

Kaiser et al. Infect Cont Hosp Epi. 2010.

WHY?

McBride et al. Applied and Env Micro 1977.

WHY?

• Temperature associated with rate of horizontal

gene transfer (NDM-1 in Delhi).

Walsh et al. Lancet ID. 2011.

WHY?

• Temperature potent driver of growth (environment).

Ratkowsky et al. J Bacteriology. 1982.

Additional Evidence

Collignon et al. Lancet Planetary Health 2018.

• 103 countries, 6 major areas (abx usage, governance,

health expenditure, GDP, education, infrastructure, climate)

• Strong significant positive univariate correlation between

temperature and resistance (E.coli and E.coli/Kleb/Staph).

Additional Evidence

Collignon et al. Lancet Planetary Health 2018.

• Significant factors in multivariable model:
• Governance index
• GDP/capita
• Infrastructure index
• Pitfalls:
• Indexes were means of standardized variables
• Variables that were averaged didn’t necessarily move in

parallel directions (e.g. temperature and precipitation)

• Why wasn’t usage significant? (positive control)

Additional Evidence

Goldstein et al. In submission.

• Longitudinal state-level resistance prevalence

data across the United States.

• NHSN (CAUTI) and IMS Quintiles state-level

antibiotic consumption.

• Mixed effects multivariable model.
• Significant positive temperature associations for

common antibiotics for E.coli and Klebsiella spp.

A warming planet with increasing population density may be further driving increases in AMR.

nasa.gov

Estimating the Burden of AMR

O’Neill. UK-AMR Review. https://amr-review.org/

Estimating the Burden of AMR

• Antimicrobial resistance is not antibiotic resistance.
• AMR - Viruses, Protozoa, Mycobacteria, Bacteria

Estimating the Burden of AMR

• Antimicrobial resistance is not antibiotic resistance.
• What is the burden of infection/pathogen?
• Many pathogens, many syndromes
• Existing surveillance not well equipped for many

community and some nosocomial infections

Estimating the Burden of AMR

• Antimicrobial resistance is not antibiotic resistance.
• What is the burden of infection/pathogen?
• What is the burden across regions/space?
• Variability of pathogen, syndromes, and prevalence of

resistance across geography.

Estimating the Burden of AMR

• Antimicrobial resistance is not antibiotic resistance.
• What is the burden of infection/pathogen?
• What is the burden across regions/space?
• What is the excess mortality/morbidity/cost?
• Many studies are underpowered
• How do you incorporate the ‘future’ costs of resistance

Estimating the Burden of AMR

• Antimicrobial resistance is not antibiotic resistance.
• What is the burden of infection/pathogen?
• What is the burden across regions/space?
• What is the excess mortality/morbidity/cost?
• Contentious.

Estimating the Burden of AMR

Estimating the Burden of AMR

Increased mortality

Estimating the Burden of AMR

Increased mortality

Estimating the Burden of AMR

Excess Length of Stay (Days)

MRSA VRE GN

Estimating the Burden of AMR

Abat et al. Clin Infect Dis 2017.

Estimating the Burden of AMR

Abat et al. Clin Infect Dis 2017.

0 to 10 million deaths annually

• Age Group
• Organism
• Syndrome

Estimating the Burden of AMR

• Argued there is NO burden of AMR in their setting.
• Little empiric data. Mostly ‘mathematical models’.
• Seek to identify XDR/MDR pathogens causing

death, as opposed to the contribution of resistance to a single or combination of drugs.

• The right approach is likely somewhere in the

middle, empirical data and models to fill the gaps.

Estimating the Burden of AMR

Estimating the Burden of AMR

• 1. Measure the burden of Sepsis.
• Assuming sepsis is an intermediate to death.
• 2. Evaluated distribution of resistant organisms

across human samples.

• 3. Evaluate relative case-fatality rates in sepsis due to

resistant vs. non-resistant.

Estimating the Burden of AMR

Challenges

• ‘Sepsis’ is increasingly unreliable diagnosis using

administrative data. ?sensitive. NOT specific.

• Not all mortality due to AMR is through ‘sepsis’.
• The distribution of resistance in clinical samples

doesn’t necessarily relate to the distribution of disease (strain related virulence, non-sterile samples).

Estimating the Burden of AMR

Another Way?

• Remove ‘sepsis’ from the equation.
• Determine marginal mortality associated with drug

resistance.

• CARI’s (Clinical Antibiotic Resistance Indices) -

proportion of mortality attributable to resistance to a given antibiotic/combination.

• Context specific. Doesn’t disentangle resistance

versus virulence. Does this matter?

Estimating the Burden of AMR

CARIs

• Use an large provincial administrative database

linked with microbiologic outcomes (Ontario).

• Marginal standardization model for hospitalization

and mortality (30 and 90 days).

• Can estimate the proportion of outcome that could

be averted by having a susceptible isolate.

• E.g. a CARI-30DM of 5% will mean that 5% of

deaths were attributable to antibiotic resistance.

Estimating the Burden of AMR

Considering Particular Antibiotic Classes

• Doesn’t make sense to use individual classes in a

binary fashion - large numbers of interactions and less clinically meaningful, only certain abx tested.

• Rather, employ a ‘generalized concentration

addition’ approach, with monotonic increasing index.

• Base this on 1st line, numbers resistant to 1st line,

then second line, etc. (Difficult-to-treat resistance)

Kadri et al. Clin Infect Dis 2018.

Objectives Summary

• 1. Antibiotic use, population density, and climate, appear to

be important predictors of the distribution of antibiotic

• resistance. Economic and hygiene related factors almost

certainly play a role - but challenging to tease these apart. 


• 2. Our understanding of the global burden of AMR is poor. It

will take years (?decades) to generate reasonable

• estimates. Thus challenging to measure progress,
• forecast. Mostly hand waving for the time being.

Acknowledgements

• David Fisman, John Brownstein, Mauricio

Santillana, Sarah McGough, Yuki Ara, Jeff Andre, Warren McIsaac, Kevin Brown, Kevin Schwartz, Nick Daneman, Gary Garber

Questions?