What is the link between Antibiotic Use and Resistance?

Herman Goossens

Correlation Between Macrolides Use and Prevalence of Erythromycin-resistant S. pneumoniae

Erythromycin-resistant S. pneumoniae (%) 60
FR

50 40 30
DE LU ESBE IT

GR

20 10
SE NL NO FI IE UK DK PT AT

0 0 1 2 3 4 5 6 Consumption of Macrolides (J01FA) in DID, AC 1998

Organism year of isolation [source of information] S. pneumoniae 1999-2000 [8]

Antibiotic resistance

Antibiotic use ATC group (year of data) Macrolides J01FA (1998)

No. of countries

Spearman correlation (r) (confidence interval) 0.83 (0.67-0.94)

P-value

Erythromycin

16

<0.001

Goossens H, et al. Lancet. 2005; 365: 579-587.

Temporal Changes in Proportion of Macrolideresistant Oral Streptococci

100 % Macrolide resistant 80 60 40 20 0 0 4 8 14 28 Days 42 180 Azithromycin Clarithromycin Placebo-1 Placebo-2

MAIN FINDINGS
Mean Use In

pre-antibiotic (Day 0) carriage of macrolide-resistant streptococci was 28% of both macrolides resulted in a huge increase in resistant streptococci, which persisted for at least 6 months (P0.01) the azithromycin group, resistance remained at a higher level than in the clarithromycin group during mid-time points (P0.001)
Malhotra-Kumar S, et al. Lancet. 2007; 369: 482-490.

Why Proving the Obvious?

As in all similar Darwinian selection systems, it is obvious that antibiotics generate resistance. It is also obvious to most that if more antibiotics are used, resistance will be more prevalent. So why prove the obvious? The answer is that the obvious correlation is not at all obvious when considered carefully.

Editorial: Antibiotic Use and Resistance - Proving the Obvious. John Turnidge and Keryn Christiansen Lancet 2005, 365: 548-9

Many Confounders in Studying the Link between Antibiotic Use and Resistance

How to measure antibiotic use (and link it with resistance)?

Which indicator? How to correct for combination therapy? How to assess dose-effect relationship? Which indicator? Colonization or positive clinical cultures? Selection bias (e.g. wrong control group)? Insufficient power? Violation of independence assumption (transmission!)?

How to measure antibiotic resistance (and link it with use)?

Which statistical methods to correlate use and resistance to address:

Outpatient Antibiotic Use in Belgium

Packages per 1,000 inhabitants per day 1997 2007, July to June

4
-1.0% -3.4% -6.4% -9.1%

3.5

-36%

Packages/1000 inh./day Verpakkingen per 1,000 inwoners per dag

-6.9% -7.5% -3.8% -3.6% -5.3%

2.5

1.5

Other J01 classes Sulfonamides and trimethoprim (J01E) Quinolones (J01M) Macrolides, lincosamides and streptogramins (J01F) Tetracyclines (J01A) Cephalosporins and other beta-lactams (J01D) Penicillins (J01C)

0.5

0 97-98 98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06 06-07

Goossens H. et al. Eurosurveillance 2008; 13:10 -13

Outpatient Antibiotic Use in Belgium

Defined Daily Doses per 1,000 Inhabitants per day 1997 2006, July to June
30

25

20

15

10

Ot h er J 0 1 classes Su lfon am id es an d t rim et h op rim (J 0 1 E) Qu in olon es (J 0 1 M) Macrolid es, lincosam id es an d st rep t og ram in s Tet racy clin es (J 0 1 A) Cep h alosp orin s an d ot h er b et a-lact am s (J 0 1 D) Pen icillin s (J 0 1 C)

DDD p er 1 0 0 0 in h . p er d ay

0 97-98 98-99 99-00 00-01 01-02 02-03 03-04 04-05 05-06

Talk Outline
Not covered:

Community Review on ecological studies linking antibiotic use and resistance

Covered:

Why measure hospital antibiotic use and resistance How to measure antibiotic use in hospitals Evidence for cause-effect relationship A few confounders Conclusions and recommendations

Why Measure Hospital Antibiotic Use and Resistance?

Measurements for judgment:

to provide league tables

Measurements for ecological purposes:

to link antibiotic use and resistance

Measurements for targeting interventions:

to show where problems are

Measurements for improvements:

to show if improvements are being made

If you cannot measure it, you cannot improve it

Lord Kelvin, 1824-1907

Hospital Antibiotic Use Has Been Neglected

Those responsible for the NHS Information Technology Strategy should consider the contrast between the excellent data on GP prescribing and the lack of data on antimicrobial use in hospitals. All hospitals should install computer systems for patient-specific prescribing information at ward level
House of Lords Science and Technology Committee Report, 1998.

How to Measure Antibiotic Use in Hospitals: Which Indicators?

DOT (Days of Treatment) Percentage of patients exposed to antimicrobials Weight (g or kg or units of treatment)/year) Grams/1,000 patient days Vials/month DDD or PDD/1000 inhabitants-days DDD or PDD/100 or 1,000 patient-days DDD or PDD/100 or 1,000 administrative bed-days DDD or PDD/100 or 1,000 occupied bed-days DDD or PDD/100 or 1,000 admissions DDD or PDD/100 or 1,000 discharges DDD or PDD/month/occupied bed

DDD, Defined Daily Dose; PDD, Prescribed Daily Dose

Which numerator to study ecological pressure?

Correlation of DDD to DOT per 1,000 Patient-days for Levofloxacin at Individual Hospitals: PDD = DDD

Polk R. et al., Clin Infect Dis, 2007; 44: 664-70

Correlation of DDD to DOT per 1,000 Patient-days for Ceftriaxone at Individual Hospitals: PDD < DDD

Polk R. et al., Clin Infect Dis, 2007; 44: 664-70

Correlation of DDD to DOT per 1,000 Patient-days for Ampicillin-Sulbactam at Individual Hospitals: PDD > DDD

Polk R. et al., Clin Infect Dis, 2007; 44: 664-70

DDD versus DOT: Reflecting Different Ecological Pressure

5 patients admitted to ward X Treated with antibiotic Y (DDD = 1 g)

Situation A:

1/5 patients treated with 3 g/daily/10 days Aggregated number of DDD: 30 Aggregated number of DOT: 10

Situation B:

3/5 patients treated with 1 g/daily/10 days Aggregated number of DDD: 30 Aggregated number of DOT: 30

Correlation between Aggregate Gentamicin Use and Resistance among Gram-negatieve Bacilli Isolates, St Pieters Hospital, Brussels, 1979-1986

% Gentamicin-resisitant Gram-negative Bacilli

Gentamicin use same year (g/year)

Gentamicin use previous year (g/year)

Goossens H, et al. Lancet 1986;2: 804.

Cause-effect Relationship: Beneath Superficiality there is Complexity

Cause:

antibiotic use infection control practises prevalence of resistance genodiversity of clones

Effect:

CAUSE

Modelling the impact of antibiotic use and infection control practices on the incidence of hospital-acquired methicillinresistant Staphylococcus aureus: a time series analysis

Aldeyab et al , J Antimicrob Chemother 2008; 62: 593-600

Estimated Multivariate Time-Series Analysis Model for Monthly HA-MRSA Incidence (R2 = 0.784)
Term
Fluoroquinolone use (DDD/100 bed-days) Third-generation cephalosporin use (DDD/100 bed-days) Macrolide use (DDD/100 bed-days) Amoxicillin/clavulanic acid use (DDD/100 bed-days) Alcohol-based handrub bulk orders Alhohol-impregnated wipes (no./100 bed-days) Patients actively screened for MRSA (no./100 bed-days) Patients admitted with MRSA (no./100 bed-days) Autoregressive term (MRSA) Moving average term (MRSA)

Time lag
1 2 4 1 3 4 2 3 2 4 2

Coefficient (SE)
0.00481 (0.00098) 0.0273 (0.00449) 0.00212 (0.00099) 0.00349 (0.000651) -0.0390 (0.0149) -0.0755 (0.0153) -0.000345 (0.0000496) -0.00721 (0.00306) 0.223 (0.0312 -0.552 (0.130) -0.980 (0.000709)

T ratio
4.905 6.080 2.149 5.365 -2.619 -4.932 -6.956 -2.357 7.162 -4.250 -1382.67

P value
<0.0001 <0.0001 0.0376 <0.001 0.0123 <0.0001 <0.0001 0.0233 <0.0001 0.001 <0.0001

Aldeyab et al, J Antimicrob Chemother 2008; 62: 593-600

EFFECT

Genodiversity of resistant Pseudomonas aeruginosa isolates in relation to antimicrobial usage density and resistance rates in intensive care units

Jonas et al, Infect Control Hosp Epidemiol 2008; 29: 350 - 7

Imipenem Usage Density and Resistance Rate in ICUs with Low (circles) and High (triangles) Genodiversity of Resistant Strains

Jonas et al, Infect Control Hosp Epidemiol 2008; 29: 350 - 7

Confounder: Colonisation Pressure

Bonten et al. Arch Intern Med 2008; 158: 1127-1132

Confounder: Underlying Illness and Length of Stay

MRSA cases OR (95% Cl) Primary covariates Levofloxacin Ciprofloxacin Other covariates Lung disease Renal disease Penicillin Metronidazole ICU stay 3.94 (2.43 to 6.40) * * 1.92 (1.10 to 3.37) 5.33 (3.28 to 8.68) 0.02 <0.001 0.02 <0.001 3.38 (1.94 to 5.90) 2.48 (1.32 to 4.67) <0.001 0.005 p value

MSSA cases OR (95% Cl) p value

0.69 (0.34 to 1.40) 0.47 (0.21 to 1.02)

0.30 0.06

2.33 (1.43 to 3.81) 1.98 (1.03 to 3.80) 1.78 (0.93 to 3.39) 1.29 (0.65 to 2.56) 4.60 (2.90 to 7.30) 1.90 (1.17 to 3.08 )

<0.001 0.04 0.08 0.46 <0.001 0.01

Emergent admission 1.74 (1.09 to 2.78)

Weber et al., Emerg Infect Dis 2003; 9: 1415-22

Confounder: Virulence
Adhesion to fibronection sigB Null S. aureus grown in either the absene or presence of 4 g of ciprofloxacin

Li et al., Antimicrob Agents Chemother. 2005; 49: 916-24

Confounder: Duration of Treatment

Objective: Estimate the probability of carbapenem resistance among P. aeruginosa isolates Method: Retrospective, cross-sectional study Adult inpatients with respiratory tract infection Log-binomial regression Results: Independent predictors of carbapenem resistance were
prior receipt of mechanical ventilation for 11 days or more prior exposure to fluoroquinolones and to carbapenems for 3 days or more

Lodisi et al., ICHE 2007; 28: 959 - 65

Confounder: Community Use

MacDougall et al. Clin Infect Dis 2005; 41: 435-440

CONCLUSIONS and RECOMMENDATIONS

Both patient-level and ecological studies of antibiotic use and resistance confirm causal relationship No agreement on which methods of measurement of antibiotic use and resistance correlate best with selection of resistance Limited understanding of host and bacterial interaction to link antibiotic exposure with emergence of resistance (host factors, clonality, fitness, virulence, pathogenicity) Pk/Pd studies are needed to investigate association of antibiotic exposure with acquisition or progression from colonisation towards infection (resistance mechanism, interaction, duration of treatment, dose, time effect) Prospective human studies are needed to better define and quantify the risks associated with antibiotic exposure