Human Health Risk Assessment of Penicillin/ Aminopenicillin Resistance in Enterococci Due to Penicillin Use in Food Animals

Risk Analysis, Vol. 29, No. 6, 2009 DOI: 10.1111/j.1539-6924.2009.01202.x

∗ , Louis Anthony (Tony) Cox, Jr.,1,2, Douglas A. Popken,1 3 and Jeremy J. Mathers4

Penicillin and ampicillin drugs are approved for use in food animals in the United States to treat, control, and prevent diseases, and penicillin is approved for use to improve growth rates in pigs and poultry. This article considers the possibility that such uses might increase the incidence of ampicillin-resistant Enterococcus faecium (AREF) of animal origin in human infections, leading to increased hospitalization and mortality due to reduced response to ampicillin or penicillin. We assess the risks from continued use of penicillin-based drugs in food animals in the United States, using several assumptions to overcome current scientiﬁc uncertainties and data gaps. Multiplying the total at-risk population of intensive care unit (ICU) patients by a series of estimated factors suggests that not more than 0.04 excess mortalities per year (under conservative assumptions) to 0.14 excess mortalities per year (under very conservative assumptions) might be prevented in the whole U.S. population if current use of penicillin drugs in food animals were discontinued and if this successfully reduced the prevalence of AREF infections among ICU patients. These calculations suggest that current penicillin usage in food animals in the United States presents very low (possibly zero) human health risks.

KEY WORDS: Animal antimicrobials; enterococcus; penicillin; resistance; risk assessment

1. INTRODUCTION coccal infections from nonhuman sources, leading to increased morbidity and mortality (WHO, 2005, Penicillin drugs (speciﬁcally, Penicillin G, Pro- 2007; FAO/WHO/OIE, 2008). Approved feed usages caine subtype) are approved for use in food animals of penicillins in food animals have been a controver- in the United States to treat, control, and prevent sial topic for several decades in the United States, diseases and to improve growth rates (FDA-CVM, with many reviews and initiatives undertaken to ad- 2007; Sechen, 2006; AHI, 2006). Concerns have dress resistance concerns (IOM, 1989; FDA, 2000; been expressed that penicillin use might increase FDA, 2003). Regulatory approvals have been given the risk of antibiotic resistance in human entero- for administration to animals via drinking water and feeds for therapeutic uses, including disease preven- tion and control, as well as for feed efﬁciency/growth 1 Cox Associates. promotion uses in food animals (FDA-CVM, 2007). 2 University of Colorado. The amounts of penicillins sold for use in food an- 3 Systems View. imals in the United States remain relatively low 4 Alpharma Animal Health Division. ∗ Address correspondence to Tony Cox, 503 Franklin Street, compared to other antimicrobials, with sole growth Denver, CO 80218, USA; tel: 303-388-1778; fax: 303-388-0609; promotion uses representing a small subfraction of [email protected]. the total (AHI, 2006). Penicillin drugs have been

796 0272-4332/09/0100-0796$22.00/1 C 2009 Society for Risk Analysis Human Health Risk Assessment of Penicillin/Aminopenicillin Resistance 797 identified as “critically important” to animal health from food animals and almost nonexistent in E. (FAO/WHO/OIE, 2008). faecalis isolates from retail meats (Hayes et al., The only penicillin drug type approved for non- 2003, 2004; McGowan et al., 2006; NARMS, 2005); therapeutic uses in the United States is Penicillin none of the meat-borne E faecalis samples isolated G (Procaine subtype), which is given primarily to from retail meats in a national surveillance program swine and chicken; none is given to cattle. A USDA (NARMS, 2002–2005) exhibited resistance to peni- (2002) survey indicated that, in grower/finisher pigs, cillin. Furthermore, ampicillin remains highly effec- over 99.5% of all penicillin is administered (primar- tive against clinical E. faecalis (Jones et al., 2004). ily via injection) for disease treatment and preven- Therefore, our risk assessment focuses on the iden- tion. A relatively small amount of ampicillin is also tified hazard of ampicillin resistance among human used for disease treatment. In a large-scale study of E. faecium infections. 82 poultry farms in the eastern United States, Hayes Risk to human health arises because some strains et al. (2004) measured resistance levels of entero- of enterococci may become opportunistic pathogens, cocci found in chicken litter. E. faecalis represented potentially resistant to multiple drugs, that infect pa- 53% of the isolates and had 0% penicillin resistance. tients who are already seriously ill (typically in ICUs) E. faecium represented 31% of the isolates. Seventy- with immune systems weakened by organ trans- one percent of the E. faecium (71 out of 104 isolates) plants, chemotherapy, AIDS, or other causes. In- had penicillin resistance, but only 1% (1 out of 104) deed, enterococcal infection is the second most com- had ampicillin resistance. mon hospital-acquired infection in the United States This article quantifies the potential for continued (Varman et al., 2006). These infections can prolong use of penicillin drugs in food animals to harm hu- illness and increase patient mortality. Vancomycin- man health by increasing the number of ampicillin- resistant enterococci (VRE) are of particular con- resistant enterococcal infections in human patients. cern because of their virulence and resistance to even After summarizing relevant background for the haz- some recently developed antibiotics. Vancomycin- ard of greatest concern—infection of intensive care resistant E. faecium (VREF) can cause serious and unit (ICU) patients with ampicillin-resistant Ente- often fatal disease in vulnerable populations, such as rococcus faecium (AREF) bacteria—the following liver transplant patients and patients with hemato- sections focus on quantifying the fraction of such re- logic malignancies (Leavis et al., 2003; Rice, 2001). sistant infections that might be prevented by discon- Many enterococcal infections, including VRE, tinuing the use of penicillin drugs in food animals. resolve without antimicrobial treatment (Varman et al., 2006; Rice, 2001). In severe cases for which antimicrobial treatment is provided, peni- 2. BACKGROUND, HAZARD cillin and ampicillin are often the leading choices. IDENTIFICATION, AND SCOPE: Other drugs that are also effective against hu- REDUCING AMPICILLIN-RESISTANT man enterococcal infections include gentamicin, E. FAECIUM INFECTIONS IN vancomycin, quinupristin-dalfopristin (Synercid R ), ICU PATIENTS linezolid (Zyvox R ), tigecycline (Tygacil R ), and ni- Enterococci are commensal gram-positive bac- trofurantion. These can be used in patients with al- teria found in the intestinal flora of most healthy lergy or high-level resistance to penicillin and ampi- birds and mammals, including people (Biavasco cillin. Many ampicillin-resistant cases can also be et al., 2007; Willems, 2000). In humans, entero- treated successfully with high doses of ampicillin, cocci typically comprise not more than 1% of the either alone or in combination with drugs such as microflora of adults (FDA-CVM, 2004) and are gentamicin or streptomycin (Varman et al., 2006; normally harmless; however, they can cause oppor- Murray, 2000). tunistic infections in ICU patients or others with Most E. faecium infections in ICU patients in weakened immune systems. The Enterococcus genus the United States are now resistant to vancomycin has 17 species, but most human clinical isolates are (Edmond et al., 1999; Jones et al., 2004; Leavis either E. faecalis (74–90%) or E. faecium (5–16%) et al., 2003). Patients with VREF have worse (Varman et al., 2006). outcomes than those with vancomycin susceptible E. faecalis infections are responsible for most strains—longer hospital stays and higher mortality clinical enterococcal infections, but penicillin (or (Webb et al., 2001). As noted by Rice (2001), vir- ampicillin) resistance is rare in E. faecalis isolates tually all VREF are also ampicillin resistant: “More 798 Cox, Popken, and Mathers than 95% of VRE recovered in the United States tions hold: (1) an ICU patient dies, following (2) an are E. faecium; virtually all are resistant to high lev- E. faecium infection that (3) is resistant to ampi- els of ampicillin.” Hence, our risk assessment treats cillin (AREF) (and hence might have benefited had VREF as being (at least approximately) a subset of ampicillin resistance been prevented). The infection AREF. One suggested explanation is “that the close was (4) vancomycin-susceptible (and hence might association of the vancomycin and ampicillin resis- have also been ampicillin-susceptible, had it not been tance phenotypes, at least in VanB-type VRE, is ex- for penicillin use in food animals); (5) not known plainable by their inclusion within large, transferable to have been contracted from the hospital environ- genetic elements” (Hanrahan et al., 2000, p. 1350). ment (and hence might have been prevented by ac- However, widespread ampicillin resistance appeared tions external to the hospital, such as elimination of in 1982 (Fortun et al., 2002), while vancomycin re- AREFs from food animals); (6) could have come sistance appeared in the early 1990s in E. faecium from food animals (i.e., has a genotype or resistance (Murray, 2000). determinants of the types found in food animals). Since most VREF are AREF (although many (7) The patient tolerated penicillin (i.e., was not aller- AREF are not VREF), and assuming that changes gic, and hence might have benefited from ampicillin, in animal penicillin use would not significantly af- had it not been for resistance). We propose that the fect vancomycin resistance (consistent with histori- conjunction of these seven conditions should be in- cal data), we focus on human (ICU patient) infec- terpreted as necessary for a mortality to have been tions with vancomycin-susceptible strains of AREF. caused (with nonnegligible probability) by resistance Presumably, this is the subpopulation that might ex- due to use of penicillin in food animals, even though perience decreased ampicillin resistance if discontin- it is not sufficient (e.g., the infecting strain might have uing animal penicillin drugs were to replace some had some other origin than food animals, or the pa- AREF cases with ampicillin-susceptible cases. For tient might have died anyway, even if the infection patients with VREF, we assume that AREF would had been ampicillin-susceptible). persist (due to the observed co-occurrence of AREF Accordingly, the following sections estimate a in VREF strains), so that no benefit from reduced plausible upper bound on annual preventable mor- AREF would be achieved for these patients. talities from AREF infections based on the following The following sections seek to quantify poten- product of factors: tially preventable AREF cases and the human health Preventable AREF mortalities per year ≤ (to- benefits that might be created if these AREF cases tal number of ICU infections per year) × (fraction were prevented (made ampicillin-susceptible) by dis- caused by E. faecium) × (fraction of ICU E. fae- continuing penicillin drug uses in food animals. This cium infections that are AREF and exogenous, i.e., approach draws on recent advances in sequencing not known to be of nosocomial origin) × (fraction technology that enable more precise strain groupings of these exogenous AREF cases that are vancomycin- and epidemiological analyses than have previously susceptible) × (fraction of vancomycin-susceptible ex- been possible. ogenous AREF cases that might have come from food animals) × (fraction of these cases that are penicillin- 3. METHODS AND DATA: UPPER BOUNDS tolerant)× (excess mortality rate for AREF cases com- FOR PREVENTABLE MORTALITIES pared to ASEF cases). Recognizing that a farm-to-fork (forward chain- That is, we first quantify the expected annual number ing) approach is not practical for AREF due to data of AREF cases in the United States that might ben- and knowledge gaps in release, exposure, and dose- efit from ampicillin treatment if food animal uses of response relations, we instead start with more readily penicillin were halted (i.e., cases that are penicillin- available human data on ICU case loads and resis- tolerant and vancomycin-susceptible and that might tance rates, similar to the approach in Cox and Pop- have been caused by resistance determinants from ken (2004). We then work backward to estimate a food animals). Then, we multiply this number by the plausible upper bound on the annual number of hu- excess mortality rate for resistant as opposed to sus- man patient mortalities that might be prevented by ceptible cases. discontinuing penicillin use in food animals. 3.1. Estimated Number of ICU Infections Per Year For purposes of conservative (i.e., upper-bound) risk assessment, we define a potentially preventable Enterococcal infections are generally limited mortality to occur whenever the following condi- to already hospitalized individuals. E. faecium Human Health Risk Assessment of Penicillin/Aminopenicillin Resistance 799 infections are frequently associated with nosoco- Kuhn et al. (2005), in Europe “it seems that animal- mial bloodstream infections occurring within ICUs. associated VRE probably reflect the former use of A recent FDA risk assessment for virginiamycin avoparcin in animal production, whereas VRE in (FDA-CVM, 2004) provided the following two ap- human-associated samples may be a result of antibi- proximate estimates: otic use in hospitals.” Since nosocomial transmission is a hospital-specific problem that can often be elim- (1) N = annual number of ICU infections = inated by rigorous control measures, we restrict our 104,372.5 based on blood-stream infections. risk assessment to exogenous (nonnosocomial) cases (2) N = 315,000 based on septicemia cases. that are potentially attributable to food animals. (If The study does not weight these alternatives. To be this restriction is dropped in sensitivity analysis, the conservative (i.e., to maximize estimated risks), we effect is simply to divide estimated annual impacts by will use the larger estimate, N = 315,000 cases/year. the nonnosocomial fraction of cases, which increases (Patients with severely complicated urinary tract in- them approximately sixfold.) fections (UTIs) are also sometimes treated with Cox and Popken (2004) used data from several intravenous antibiotics, including combinations of studies in the 1990s to estimate an approximate mean gentamicin and ampicillin, but ceftriaxone may be value of 0.17 for the fraction of exogenous cases in substituted for ampicillin if needed, and oral an- the United States. (This is the mean of a range of val- tibiotics (e.g., trimethoprim, cephalosporins, nitro- ues, extending from a low of 0.089 based on data in furantoin, or ciprofloxacin) are used in the vast Bischoff et al. (1999) to a high of 25% based on esti- majority of cases (http://adam.about.com/reports/ mates in Austin et al. (1999).) More recent investiga- 000036 7.htm). We therefore do not include UTI tions of the molecular epidemiology of drug-resistant cases in this assessment.) E. faecium infections suggest that, if anything, this proportion may have decreased since the 1990s as a particular hospital-adapted clone of E. faecium called 3.2. Fraction of ICU Infections Caused CC17 has spread widely in hospitals in the United by E. faecium States and elsewhere (Leavis et al., 2006; Top et al., The proportion of ICU infections that are caused 2007; Willems, 2001). Approximately 88% of E. fae- by E. faecium can be estimated with the help of the cium isolates from hospital outbreaks (n = 32) be- following two fractions from the same FDA-CVM long to Complex-17, compared to 59% of all clini- study: cal isolates (n = 162), 23% of isolates from hospital surveillance in Australia, Europe, and North and (1) P = 0.10 = fraction of ICU infections ent South America (n = 64), 5% of community isolates caused by Enterococcus. (Wisplinghoff et al., (n = 57), and 1% of isolates from animal surveil- 2004 provide an estimate of 0.09. To be con- lance (n = 96, including bison, calves, cats and dogs, servative, we use the higher estimate of 0.10.) ostriches, poultry, pigs, and rodents in Africa and (2) P , = 0.25 = fraction of enterococcal in- EF ent Europe) (Leavis et al., 2006). In the United States, fections caused by E. faecium. too, Complex-17 and a closely associated clade of The product of these two factors, fEF = Pent ∗ hospital-associated strains dominate the epidemiol- PEF,ent = 0.025, is the estimated fraction of ICU in- ogy of AREF (Leavis et al., 2007; Deshpande et al., fections caused by E. faecium. An approximate value 2007). for the expected annual rate of E. faecium infections To be conservative, we continue to assume that can then be obtained via the equation: the fraction of exogenous cases is 0.17, although ac- Expected annual number of E. faecium infections = knowledging that this fraction may be declining sig- N ∗ fEF = N ∗ Pent ∗ PEF,ent ≤ 315,000 ∗ 0.025 = 7,875 nificantly as these hospital-associated strains account E. faecium infections/year. for a higher proportion of all E. faecium infections. The fraction of exogenous cases that are ampicillin-resistant can be estimated from data in 3.3. Fraction of ICU E. faecium Infections that are Table I of Willems et al. (2005). Pooling all nonout- Ampicillin-Resistant and Exogenous break, non-Complex-17 cases (where “nonoutbreak” (Nonnosocomial) cases are the sum of clinical and hospital surveil- Most AREF infections are contracted nosoco- lance isolates) gives a total of 20 ampicillin-resistant mially. Indeed, it is possible that few or none orig- cases out of 107 total cases, for a resistance frac- inate in food animals. For example, as stated by tion of: 20/107 = 0.187 ampicillin-resistant cases 800 Cox, Popken, and Mathers per nonoutbreak case. In summary, the estimated multiple strains of E. faecium. Each number expected annual number of ampicillin-resistant, represents a sequence type (ST). Lines connect STs exogenously caused (i.e., nonnosocomial) E. faecium that differ in only one of seven “housekeeping infections in the United States is no more than: genes.” The relative sizes of the circles represent the relative prevalences of the STs. In the relatively few (7,875 E. faecium infections/year) ∗ (0.17 nonnosocomial fraction) ∗ (0.187 ampicillin-resistant fraction) = cases where human patient and food animal (pig and 7,875 ∗ 0.17 ∗ 0.187 = 250 exogenous AREF infections poultry) clusters overlap, the strains falling in the per year. overlap might have come from a common environmental source (e.g., soil or water), or might be due 3.4. Fraction of Vancomycin-Susceptible Cases to a “reverse causation” flow from humans to pigs via surface water, flies, pets, or other paths (Guard- Assuming that almost all vancomycin-resistant abassi & Dalsgaard, 2004; Macovei & Zurek, 2006; strains of E. faecium in the United States are also Rodrigues et al., 2002). ampicillin resistant (but not vice-versa) (Rice, 2001), The Multi Locus Sequence Typing website the relatively recent data of Jones et al. (2004) show (www.mlst.net—curator: Rob Willems) provides a that, in the United States, about 14% of E. faecium database of 490 E. faecium samples—a subset of strains are ampicillin resistant and vancomycin sus- those used to generate Fig. 1 of Leavis et al. ceptible. Specifically, 90.3% of E. faecium isolates (2006). The data indicate 87 unique STs among were resistant to ampicillin and 76.3% of E. fae- “Clin˙Isol” and “Hosp˙Surv” (human, nonoutbreak, cium isolates were resistant to vancomycin. The dif- noncommunity) clusters. (We do not consider the = ference is 0.903 – 0.763 0.14. This is an estimate “Hosp Outbreaks” category, since these are assumed of the fraction of E. faecium isolates that are AREF to fall into the nosocomially transmitted group. We but not VREF—in other words, the vancomycin- also do not consider the “Human comm” category susceptible AREF of interest for our risk assessment. as these are noninfectious strains found in healthy = Thus, 0.14/0.903 0.155 is the estimated fraction of individuals.) Two of these 87 STs (26 and 32) are AREF that are vancomycin susceptible. Using this shared with poultry and four (5, 6, 18, 133) are point estimate yields: shared with pigs. If we assume, conservatively, that Expected exogenous ampicillin-resistant and all shared types represent transmission from food an- vancomycin-susceptible cases per year ≤ 250 ∗ 0.155 = imals to human patients (rather than from people 38.75 vancomycin-susceptible AREF infections/year. to animals, or to both from common environmental This should be considered an upper bound. For ex- sources such as soil, water, flies, or birds), then an ample, research by Suppola et al. (1999) suggested estimate of the fraction of strains in human patients = that Van A and Van B E. faecium incorporate into that might originate in food animals is: 6/87 0.069. an endemic vancomycin-susceptible AREF strain. This assumes that all shared types arise from foodborne transmission of resistant bacteria, originating in food animals on the farm, to people via the food 3.5. Fraction of Exogenous Cases Potentially chain. Using this point estimate reduces the above es- from Food Animals timate to: As reviewed above, genetic similarities between ampicillin-resistant strains found in nonoutbreak Expected exogenous vancomycin-susceptible and ampicillin-resistant cases per year from food animals E. faecium infections among hospitalized patients ≤ (38.75 exogenous vancomycin-susceptible and (most of which carry the esp virulence gene and other ampicillin-resistant cases per year) ∗ (fraction of not distinctive genes) and strains found in food animals more than 6/87 from food animals) ≤ 2.67 exoge- (most of which do not) is weak (Kuhn et al., 2005; nous vancomycin-susceptible and ampicillin-resistant Leavis et al., 2006, 2007). No clear empirical attribu- infections per year from food animals. tion of hospital cases to food animals can be made 3.6. Penicillin Allergies based on these data. Fig. 1 summarizes data that suggest a possible Hospitalized patients who are allergic to peni- upper-bound quantitative estimate for the contri- cillin cannot have their enterococcal infections bution of strains of E. faecium found in food ani- treated with penicillin or ampicillin. Since such pa- mals to strains found in nonoutbreak (non-Complex- tients are not harmed by penicillin resistance, we 17) human patient isolates. The figure represents need to exclude them from risk calculations. A inferred patterns of evolutionary descent among large U.S. study of hospitalized patients requiring Human Health Risk Assessment of Penicillin/Aminopenicillin Resistance 801

Fig. 1. Isolates from human hospital patients belong to a disjoint cluster from isolates in pigs and healthy human VRE, and are almost disjoint from poultry isolates in this population snapshot of 855 E. faecium isolates on the basis of MLST allelic profiles using the eBURST algorithm (Leavis et al., 2006). This snapshot shows all clonal complexes, singletons, and patterns of evolutionary descent. The relative size of the circles indicates their prevalence in the MLST database (http://www.mlst.net/). Numbers correspond to the STs, and lines connect single locus variants: STs that differ in only one of the seven housekeeping genes. CC17, the major subpopulation representing hospital outbreaks and clinical infections, is indicated, as well as the source of other major subgroups. Annotations: Clin infect, isolates from clinical sites (mainly blood) from hospitalized patients; Human comm, faeces isolates from human volunteers not connected to hospitals; Hosp outbreak, isolates from hospital outbreaks; Hosp surv, faeces isolate from hospitalized patients without an enterococcal infection and not associated with an enterococcal hospital outbreak; VSE, vancomycin-susceptible enterococci. Source: Reproduced with permission from Leavis et al. (2006). antimicrobials found that 15.6% reported an allergy between ampicillin-resistant cases and ampicillin- to penicillin (Lee et al., 2000). (This exceeds the av- susceptible cases. They stated that: erage in the general population, which is expected.) The remaining (1 – 0.156) = 0.844 of patients corre- There were no significant differences in the outcome of patients with ampicillin-resistant and -susceptible sponds to: strains. We did not find significant differences in mor- Expected exogenous vancomycin-susceptible tality between the two groups. Overall mortality in pa- ampicillin-resistant cases per year from food ani- tients with bacteraemia caused by ampicillin-resistant mals, in penicillin-tolerant patients ≤ 0.844 × 2.67 = and -susceptible E. faecium was 34% and 21%, respec- 2.25. tively (OR: 2.1; 95% CI: 0.47–9.95). Mortality attributed to bacteraemia was 21% and 15%, respectively (OR: 3.7. Excess Mortalities 1.5; 95% CI: 0.27–8.85). (Fortun et al., 2002, p. 4) The next step is to calculate the increase in hu- To obtain a nonzero risk estimate despite the re- man health harm—especially, increased mortality— ported absence of statistically significant differences among the 2.25 expected cases per year calculated in mortality, we make the conservative assump- in the previous steps. Fortun et al. (2002) reported tion that the numerical difference in bacteraemia- no statistically significant differences in outcomes attributed mortality rates between patients with 802 Cox, Popken, and Mathers ampicillin-resistant and -susceptible strains reflects tainty about the validity and conservatism of the as- a true causal effect (i.e., that resistance does cause sumptions in Table I. a 21% – 15% = 6% increase in absolute mortality Qualitatively, the main uncertainty is about risk, per patient per infection). In other words, we whether there is a nonzero risk to human health from treat the statistically nonsignificant difference as a animal use of penicillin drugs. We have assumed that true difference caused by ampicillin-resistance (but there is, but there is no clear empirical proof that the acknowledge that this is not the original authors’ in- risk is nonzero. To bridge this knowledge gap, Ta- terpretation and that the true difference could be ble I incorporates several conservative qualitative as- as small as zero). This assumption provides a possi- sumptions that jointly imply that the risk is nonzero. ble basis for calculating a nonzero human health risk Other quantitative parameter values presented, and from ampicillin resistance. their implied risk estimate of ≤0.135 excess mortal- With this assumption, the expected annual ex- ities/year, are intended to be realistic, data-driven cess mortality risk caused by ampicillin resistance values (rather than extreme upper bounds or 95% becomes: upper confidence limits) contingent on these conservative qualitative assumptions. Expected excess mortalities per year (for the entire The most important conservative elements in Ta- U.S. population) caused by exogenous vancomycin- susceptible and AREF infections, assumed to originate ble I are the following qualitative assumptions: ≤ from food animals, in penicillin-tolerant patients 2.25 (1) Transfer of ampicillin resistance from food an- ∗ 0.06 = 0.135 excess mortalities/year. imal bacteria to bacteria infecting human pa- In reality, annual mortality risks from AREF tient occurs. The assumption that ampicillin- are likely to be much smaller than this, as patient- resistant strains and/or determinants are cultured isolates would typically be screened for re- transferred from strains in food animals to hu- sistance prior to treatment (standard procedure in- man ICU patients is fundamental to the as- dicated for such infections) and then patients with sessment in Table I. Such transfer has never AREF would be treated with other drugs such been shown to occur, but may be possi- as gentamicin, vancomycin, quinupristin/dalfopristin, ble, based on the similarities described in linezolid, daptomycin, tigecycline, or nitrofuran- Fig. 1. tion. In addition, the above mortality estimate does (2) Withdrawing animal drug use would imme- not address morbidity or quality-adjusted life years diately and completely prevent the problem. (QALYs) lost due to potentially preventable re- Table I assumes that halting penicillin use sistance. The patients at risk are already severely in food animals would immediately eliminate ill (usually, immuno-compromised) patients such all ampicillin resistance from the cases in as leukemia, transplant, and AIDS patients. Thus, Table I. This is a deliberately extreme as- the hypothesized increased risk (per infection with sumption. In reality, halting use might have ampicillin-resistant vancomycin-susceptible E. fae- little or no impact on the already very low lev- cium) represents fewer QALYs lost than would els of ampicillin resistance. be the case for otherwise healthy patients. We (3) Resistance increases patients mortality. The as- have therefore not attempted to quantify QALY sumption that ampicillin resistance causes an impacts. increase in the mortality rates of the patients in Table I is made even though, in reality, no statistically significant difference in mortality 4. RESULTS SUMMARY, SENSITIVITY rates has been found between resistant and AND UNCERTAINTY ANALYSIS nonresistant cases (Fortun et al., 2002). Table I summarizes key parameter estimates, With these assumptions, the calculations in calculations, assumptions, and resulting risk esti- Table I predict that excess mortalities per year in mates from this study. It is traditional in presenting the entire U.S. population could be as high as 0.135, point estimates to also present interval estimates to i.e., an excess mortality expected roughly once ev- inform decisionmakers about the plausible range of ery seven to eight years, if current conditions per- estimated values. In the present analysis, however, sist. This risk is concentrated among ICU patients the key uncertainties have little to do with statistical already at high risk of such infections. With less sampling error, and are not adequately characterized conservative assumptions, the estimated risk falls to by confidence limits. Rather, they arise from uncer- about 0.04 excess mortalities per year, i.e., about one Human Health Risk Assessment of Penicillin/Aminopenicillin Resistance 803

Table I. Summary of Risk Calculation

Possible Factor Base Case Value Alternative Value Source

N = ICU infections/year N = 315,000 N = 104,372.5 FDA-CVM, 2004 Pent = fraction of ICU infections caused by 0.10 0.09 (Wisplinghoff et al., FDA-CVM, 2004 Enterococcus 2004) PEF,ent = fraction of enterococcal infections caused 0.25 FDA-CVM, 2004 by E. faecium. Fraction of enterococcal infections caused by ≤ 0.17 Cox and Popken, 2004. E. faecium that are exogenous (nonnosocomial) (May be smaller now due to spread of CC-17) Fraction of exogenous cases that are ampicillin 0.187 Willems et al., 2005 resistant Fraction of exogenous ampicillin-resistant cases 0.155 Jones et al., 2004 that are vancomycin susceptible Fraction of exogenous ampicillin-resistant 0–0.069 (0.069 assumed) Data of Leavis et al., 2006 vancomycin-susceptible cases possibly from food animals Fraction of exogenous ampicillin-resistant cases 0.844 Lee et al., 2000 with penicillin-tolerant host Fraction of these cases that would become 0.00–1.00. (1 is assumed) Conservative assumption ampicillin susceptible if penicillin use in food animals were terminated Increase in mortality risk per case, due to ampicillin 0.00–0.06. (0.06 is assumed) Fortun et al., 2002, resistance conservative assumption RISK ≤ 0.14 potential excess mortalities/year 315000 ∗ 0.10 ∗ 0.25 ∗ 0.17 ∗ 104372.5 ∗ 0.09 ∗ 0.25 ∗ Product of preceding 0.187 ∗ 0.155 ∗ 0.069 ∗ 0.17 ∗ 0.187 ∗ 0.155 ∗ factors 0.844 ∗ 0.06 = 0.135 0.069 ∗ 0.844 ∗ 0.06 ≈ 0.04 mortalities/year excess mortality every 25 years in the United States are violated, then the true risk could be as low as under current conditions. The multiplicative calcu- zero. lation in Table I makes sensitivity analysis of these We have emphasized that the most important results to changes in the values of specific factors uncertainty in this analysis is discrete—is the pre- especially straightforward: the final risk estimate is ventable fraction of risk positive or is it zero?—and directly proportional to each factor listed. that such uncertainty is not well characterized by a The more conservative risk estimate of 0.135 ex- confidence interval. Nonetheless, it may be useful to cess mortalities per year equates to an average in- consider a rough upper bound on how large the true dividual risk rate in the most at-risk group (ICU risk might be. A crude estimate is given by Markov’s patients) of approximately 0.135/315,000 = 4.3 × inequality for nonnegative random variables if we as- 10−7 excess mortalities per ICU patient. For the sume that the risk estimates in Table I represent ex- U.S. population as a whole, this corresponds to pected values. In this case, a (possibly extreme) up- an average individual risk of approximately 0.135/ per bound on the true but unknown risk is that it has 300E6 = 4.5 × 10−10 excess fatalities per person-year, at most a 5% probability of exceeding the point es- or a lifetime risk of about 80(6 × 10−10) = 3.6 × timates (0.135 or 0.04 excess mortalities per year) by 10−8 excess risk of mortality per lifetime (for an as- more than 20-fold. To the extent that these point esti- sumed 80-year lifetime). This is well below the risk mates are biased upward by the assumptions listed in level of 1 × 10−6 (1 per million lifetimes) sometimes Table I, upper bounds based on Markov’s inequality cited as a threshold for concern for carcinogens in the will be even more conservative. environment. If the less conservative risk estimate of 0.04 excess mortalities per year is used, these in- 5. DISCUSSION AND CONCLUSIONS dividual and population risks are reduced by a factor of 0.04/0.135, or more than threefold. If one or Concerns about penicillin use in food animals more of the key qualitative assumptions listed above and potential transfer of resistance to humans have 804 Cox, Popken, and Mathers been debated over several decades. However, cur- Enterococci from humans, animals and food: Species distri- rent knowledge and data, as analyzed in Table I, sug- bution, population structure, Tn1546-typing and location, and virulence determinants. Applied and Environmental Microbi- gest that ongoing penicillin usage in food animals in ology, 2007; 73(10):3307–3319. the United States creates at most quite minor mortal- Bischoff WE, Reynolds TM, Hall GO, Wenzel RP, Edmond MB ity risks to human health. Quantitatively, it appears Molecular epidemiology of vancomycin resistant Enterococ- cus faecium in a large urban hospital over a 5-year period. that these risks are unlikely to exceed one potentially Journal of Clinical Microbiology, 1999; 37(12):3912–3916. preventable mortality in the U.S. population roughly Cox LA, Popken DA. Quantifying human health risks from vir- every 7–25 years. The true value could be smaller giniamycin used in chickens. Risk Analysis, 2004; 24(1):271– 288. (and is zero if any of our key conservative qualitative Deshpande LM, Fritsche TR, Moet GJ, Biedenbach DJ, Jones assumptions are incorrect). RN. Antimicrobial resistance and molecular epidemiology of Removing penicillin drugs from the animal drug vancomycin-resistant enterococci from North America and Europe: A report from the SENTRY antimicrobial surveil- market is a possible risk management option (and lance program. Diagnostic Microbiology and Infectious Dis- has actually been proposed). But significant base- ease, 2007; 58(2):163–170. line resistance among antibiotic-free animals (Patton Edmond MB, Wallace SE, McClish DK, Pfaller M A, Jones RN, Wenzel RP. Nosocomial bloodstream infections in United et al., 2006), along with small potential human health States hospitals: A three-year analysis. Clinical Infectious Dis- benefits even if removal immediately eliminated eases, 1999; 29(2):239–244. all preventable cases of resistance in human pa- FAO/WHO/OIE. Joint FAO/WHO/OIE Expert Meeting on Crit- ically Important Antimicrobials. Report of a meeting held in tients (as Table I assumes) suggests that even com- FAO, Rome, Italy, 26–30 November 2007. FAO, Rome, Italy, plete product removal would not detectably improve and WHO, Geneva, Switzerland, 2008. human health. Increased surveillance of food- Fortun J, Coque TM, Martin-Davila P, Moreno L, Canton R, Loza E, Baquero F Moreno S. Risk factors associated with ampi- animal-associated enterococci and tracking their cillin resistance in patients with bacteraemia caused by En- penicillin and ampicillin resistances (NARMS, 2005), terococcus faecium. Journal of Antimicrobial Chemotherapy, as well as compliance with judicious use guidelines 2002; 50(6):1003–1009. Guardabassi L, Dalsgaard A. Occurrence, structure, and mo- for practitioners and producers (AVMA, 2008), may bility of Tn1546-like elements in environmental isolates of suffice to protect human health against the current vancomycin-resistant enterococci. Applied and Environmen- small risks without compromising the health of food tal Microbiology, 2004; 70(2):984–990. Hanrahan J, Hoyen C, Rice LB. Geographic distribution of a large animals. mobile element that transfers ampicillin and vancomycin resistance between Enterococcus faecium strains. Antimicrobial Agents and Chemotherapy, 2000; 44:1349–1351. ACKNOWLEDGMENTS Hayes JR, English LL, Carr LE, Wagner DD Joseph SW Multiple- antibiotic resistance of Enterococcus spp. isolated from com- We thank Alpharma for supporting the re- mercial poultry production environments. Applied and Envi- search reported here. The research questions ad- ronmental Microbiology, 2004; 70(10):6005–6011. Hayes JR, English LL, Carter PJ, Proescholdt T, Lee KY, Wagner dressed, methods and models applied, and conclu- DD, White DG. Prevalence and antimicrobial resistance of sions reached are solely those of the authors. Enterococcus species isolated from retail meats. Applied and Environmental Microbiology, 2003; 69(12):7153–7160. Institute of Medicine (IOM). Human Health Risks with the Sub- REFERENCES therapeutic Use of Penicillin or Tetracyclines in Animal Feed. Report by the Committee of Human Health Risk Assess- American Veterinary Medical Association (AVMA). American ment of Using Subtherapeutic Antibiotics in Animal Feeds, Veterinary Medical Association Judicious Therapeutic Use Institute of Medicine, IOM-88-89. Washington, DC: National of Antimicrobials December, 2005. Available at http://www. Academy Press, 1989. avma.org/issues/jtua/default.asp, 2008 (accessed December Jones ME, Draghi DC, Thornsberry C, Karlowsky JA, Sahm DF, 31, 2008). Wenzel RP Emerging resistance among bacterial pathogens Animal Health Institute (AHI). Animal Health Companies Meet in the intensive care unit—A European and North Amer- Increase Market Need for Antibiotics. Washington, DC: ican surveillance study (2000–2002). Annals of Clinical News Release from the Animal Health Institute, Octo- Microbiology and Antimicrobials, 2004; 3(14): 1–11. Avail- ber 12, 2006. Available at http://www.ahi.org/mediacenter/ able at http://www.pubmedcentral.nih.gov/articlerender.fcgi? documents/Antibioticuse2005REVISED.pdf (accessed De- artid=509280 (accessed December 31, 2008). cember 31, 2008). Kuhn I, Iversen A, Finn M, Greko C, Burman LG, Blanch AR, Austin DJ, Bonten MJM, Weinstein RA, Slaughter S, Anderson Vilanova X, Manero A, Taylor H, Caplin J, Dominguez L, RM. Vancomycin-resistant enterococci in intensive-care hos- Herrero IA, Moreno MA, Mollby R. Occurrence and relat- pital settings: Transmission dynamics, persistence, and the im- edness of vancomycin-resistant enterococci in animals, hu- pact of infection control programs. Proceedings of National mans, and the environment in different European regions. Academy of Sciences of USA, 1999; 96:6908–6913. Available Applied and Environmental Microbiology, 2005; 71(9):5383– at http://www.pnas.org/ (accessed December 31, 2008). 5390. Biavasco F, Foglia G, Paoletti C, Zandri G, Magi G, Guaglianone Leavis HL, Bonten MJ. Willems RJ. Identification of high-risk en- E, Sundsfjord A, Pruzzo C, Donelli F, Facinelli B. VanA-type terococcal clonal complexes; global dispersion and antibiotic Human Health Risk Assessment of Penicillin/Aminopenicillin Resistance 805

resistance. Current Opinions in Microbiology, 2006; 9(5):454– U.S. Food and Drug Administration—Center for Veterinary 460. Medicine (FDA-CVM). Risk Assessment of Streptogramin Leavis HL, Willems RJ, van Wamel WJ, Schuren FH, Caspers MP, Resistance in Enterococcus Faecium Attributable to the Bonten MJ. Insertion sequence-driven diversification creates Use of Streptogramins in Animals. Draft for Comment, a globally dispersed emerging multiresistant subspecies of E. November 23, 2004. Available at http://www.fda.gov/cvm/ faecium. PLoS Pathogens, 2007; 3(1): e7. Documents/SREF RA FinalDraft.pdf (accessed December Leavis HL, Willems RJ, Top J, Spalburg E, Mascini EM, Fluit 31, 2008). AC, Hoepelman A, de Neeling AJ, Bonten MJ. Epidemic U.S. Food and Drug Administration—Center for Veterinary and nonepidemic multidrug-resistant Enterococcus faecium. Medicine (FDA-CVM). Guidance for Industry # 152: Eval- Emerging Infectious Diseases, 2003; 9(9); 1108–1115. uating the Safety of Antimicrobial New Animal Drugs with Lee CE, Zembower TR, Fotis MA, Postelnick MJ, Greenberger Regard to Their Microbiological Effects on Bacteria of PA, Peterson LR, Noskin GA. The incidence of antimicrobial Human Health Concern. U.S. Department of Health and allergies in hospitalized patients: Implications regarding pre- Human Services, Food and Drug Administration, Center scribing patterns and emerging bacterial resistance. Archives for Veterinary Medicine, 2003. Available at http://www.fda. of Internal Medicine, 2000; 160(18):2819–2822. gov/cvm/Guidance/fguide152.pdf (accessed December 31, Macovei L, Zurek L. Ecology of antibiotic resistance genes: Char- 2008). acterization of enterococci from houseflies collected in food U.S. Food and Drug Administration—Center for Veterinary settings. Applied and Environmental Microbiology, 2006; Medicine (FDA-CVM). Review of Agricultural Antibi- 72(6):4028–4035. otics Policies, 2000. Available at http://www.fda.gov/cvm/ McGowan LL, Jackson CR, Barrett JB, Hiott LM, Fedorka-Cray. HRESP106 157.htm#nrdc (accessed December 31, 2008). Prevalence and antimicrobal resistance of enterococci iso- Varman M, Chatterjee A, Abuhammour W, Johnson WC. lated from retail fruits, vegetables, and meats. Journal of Food Enterococcal infection, 2006. Emedicine.com. Available at Protection, 2006; 69(12): 2976–2982. http://www.emedicine.com/ped/topic2703.htm (accessed July Murray BE. Vancomycin-resistant enterococcal infections. New 26, 2006). England Journal of Medicine, 2000; 342(10):710–721. Webb M, Riley LW, Roberts RB. Cost of hospitalization for and National Antimicrobial Resistance Monitory System (NARMS). risk factors associated with vancomycin-resistant Enterococ- NARMS Retail Meat Annual Reports, 2002–2005. Avail- cus faecium infection and colonization. Clinical Infectious able at http://www.fda.gov/cvm/narms˙pg.html#Publications Diseases, 2001; 33(4):445–452. (accessed December 31, 2008). Willems RJ, Top J, van Santen M, Robinson DA, Coque TM, Patton BS, Cast WR, Kocher ME, Matthews JO, Griffith RW, Baquero F, Grundmann H, Bonten MJ. Global spread Hurd HS, Dickson JS. The Effect of Swine Production System of vancomycin-resistant Enterococcus faecium from distinct on Bacterial Prevalence and Antibiotic Resistance. Procced- nosocomial genetic complex. Emerging Infectious Diseases, ings of the Poster Presentation at International Association 2005; 11(6), 821–828. for Food Protection (IAFP) 2006, Calgary, Alberta, Canada, Willems RJ, Horman W, Top J, van Santen-Verheuvel M, August 13–16, 2006. Abstract available at http://www.food Tribe D, Manzioros X, Gaillard C, Vandenbroucke-Grauls protection.org/meetingsEducation/IAFP%202006/IAFP% CMJE, Mascini EM, van Kregten E, van Embden JDA 2006%20Poster%20Abstracts.pdf (accessed December 31, Bonten MJ. Variant esp gene as a marker of a distinct 2008). genetic lineage of vancomycin resistant Enterococcus fae- Rice LB. Emergence of vancomycin resistant enterococci. Emerg- cium spreading in hospitals. Lancet, 2001; 357(9259), 853– ing Infectious Diseases, 2001; 7(2):183–187. 855. Rodrigues J, Poeta P, Martins A, Costa D. The importance Willems RJ, Top J, Van Den Braak N, van Belkum A, Endtz H, of pets as reservoirs of resistant Enterococcus strains, with Mevius D, Stobberingh E, Van Den Bogaard A, van Emb- special reference to vancomycin. Journal of Veterinary den JDA. Host specificity of vancomycin-resistant Enterococ- Medicine B, 2002; 49(6):278–280. cus faecium. Journal of Infectious Diseases, 2000; 182(3):816– Sechen S. The review of animal production drugs by FDA. FDA 823. Veterinarian, 2006; 21(1):8–11. Available at http://www.fda. Wisplinghoff H, Bischoff T, Tallent SM, Seifert H, Wenzel RP, gov/cvm/Documents/FDAVetVolXXINo1.pdf (accessed De- Edmond MB. Nosocomial bloodstream infections in US hos- cember 31, 2008). pitals: Analysis of 24,179 cases from a prospective nation- Suppola J, Kolho E, Salmenlinna S, Tarkka E, Vuopio-Varkila wide surveillance study. Clinical Infectious Diseases, 2004; J, Vaara M. VanA and VanB incorporate into an endemic 39(3):309–317. ampicillin-resistant vancomycin-sensitive Enterococcus fae- World Health Organization (WHO). Critically Important An- cium strain: Effect on interpretation of clonality. Journal of tibacterial Agents for Human Medicine for Risk Management Clinical Microbiology, 1999; 37(12):3934–3939. Strategies of Non-Human Use. Report of a WHO Work- Top J, Willems R, Blok H, de Regt M, Jalink K, Troelstra A, ing Group Consultation February 15–18, 2005. Canberra, Goorhuis B, Bonten M. Ecological replacement of Enterococ- Australia, 2005. Available at http://www.who.int/entity/ cus faecalis by multiresistant clonal complex-17 Enterococcus foodborne disease/resistance/FBD CanberraAntibacterial faecium. Clinical Microbiology and Infection, 2007; 13(3):316– FEB2005.pdf (accessed December 31, 2008). 319. World Health Organization (WHO). Joint FAO/WHO/OIE U.S. Food and Drug Administration—Center for Veterinary Meeting on Critically Important Antimicrobials. Report of Medicine (FDA-CVM). FDA Database of Approved Ani- the FAO/WHO/OIE Expert Meeting. FAO Rome Italy. mal Drug Products. FDA Center for Veterinary Medicine, November 26–30, 2007. Available at http://www.who.int/ VMRCVM Drug Information Lab, 2007, Available at foodborne disease/resources/Report%20joint%20CIA%20 http://dil.vetmed.vt.edu/ (accessed December 31, 2008). Meeting.pdf (accessed December 31, 2008).