Antimicrobial Resistance Among Nontyphoidal Salmonella Isolated From Blood in the United States, 2003–2013

Salmonella causes an estimated 1.2 million illnesses, 23 000 hospitalizations, and 450 deaths annually in the United States [1]. Nontyphoidal Salmonella enterica causes the majority of these infections and almost all Salmonella-related hospitalizations and deaths [2]. Most Salmonella infections result from the ingestion of contaminated food [2, 3] and are characterized by gastroenteritis. Invasive infection, such as bacteremia and meningitis, occurs most commonly in persons with compromised immunity [4], including those with human immunodeficiency virus infection [5], infants [6], and older adults, who may have increased risk of complications, including death [7, 8]. Invasive nontyphoidal Salmonella infection is most common among patients with serotypes Choleraesuis, Dublin, Enteritidis, Heidelberg, Poona, and Schwarzengrund [6].

Antimicrobial treatment can be life-saving for invasive Salmonella infections [2]. Antibiotics commonly prescribed for these infections include fluoroquinolones (eg, ciprofloxacin) or third-generation (extended-spectrum) cephalosporins (eg, ceftriaxone) [9]. Antimicrobial resistance may contribute to bacteremia, treatment failure, and poor clinical outcomes. Hospitalizations occur with increased frequency in persons with resistant isolates, particularly those with ceftriaxone resistance [10, 11]. Salmonella bacteremia is more common in drug-resistant than in susceptible infections [10–12]. For example, fluoroquinolone resistance is associated with a >3-fold increased risk of invasive illness or death within 90 days, and nalidixic acid resistance may correlate with ciprofloxacin treatment failure [13, 14]. In recent years, azithromycin use for nontyphoidal Salmonella treatment has increased, probably owing to increasing resistance to fluoroquinolones and extended-spectrum cephalosporins; however, there have been recent reports of azithromycin treatment failures [15]. The Clinical and Laboratory Standards Institute (CLSI) has not yet established a break point for azithromycin resistance for nontyphoidal Salmonella [3, 16].

Most nontyphoidal Salmonella infections are food borne. Resistance among nontyphoidal Salmonella has been linked to antimicrobial use in food animal production [17, 18]. Injudicious antimicrobial use among humans has also been linked to an increased risk of antimicrobial-resistant infection [19].

We describe and compare antimicrobial resistance patterns among nontyphoidal Salmonella blood and stool isolates and trends in antimicrobial resistance. We compare characteristics (sex and age) of persons with nontyphoidal Salmonella isolated from blood versus stool. We also assess the differences in blood isolation rates and resistance by serotype.

MATERIALS AND METHODS

The National Antimicrobial Resistance Monitoring System (NARMS) at the Centers for Disease Control and Prevention (CDC) has tracked resistance patterns among enteric pathogens from humans since 1996. NARMS is a collaboration among the CDC, the US Food and Drug Administration (FDA), the US Department of Agriculture, and state and local public health departments. Since 2003, NARMS has included health departments from all 50 states, covering >300 million persons [20]. Participating public health laboratories submit every 20th nontyphoidal Salmonella isolate to the CDC laboratory and include available information regarding age, sex, specimen source, and serotype.

Isolates were tested in the CDC NARMS laboratory using broth microdilution (Sensititre; Trek Diagnostics, part of Thermo Fisher Scientific). The following antimicrobial agents were tested routinely via this method from 2003 through 2013: amoxicillin–clavulanic acid, ampicillin, cefoxitin, ceftiofur, ceftriaxone, chloramphenicol, ciprofloxacin, gentamicin, kanamycin, nalidixic acid, streptomycin, sulfisoxazole/sulfamethoxazole (sulfonamide), tetracycline, and trimethoprim-sulfamethoxazole (TMP-SMX). CLSI criteria were used for interpretation when available [21]. Azithromycin was not routinely tested before 2011. Before 2004, sulfamethoxazole was used instead of sulfisoxazole. NARMS break points were used for streptomycin (≥64 µg/mL) and azithromycin (≥32 µg/mL), which have no CLSI break points [20].

Isolates with specimen sources other than blood or stool and isolates of serotypes Typhi, Paratyphi A, Paratyphi B (var L+ tartrate−), and Paratyphi C were excluded from the analysis. Blood and stool isolates that were not serotyped, were partially serotyped, or were characterized as rough or nonmotile were also excluded. From this point forward, we refer to all nontyphoidal Salmonella isolates included in the analysis as Salmonella. All resistant isolates and isolates with an intermediate minimum inhibitory concentration were categorized as resistant to remain in accordance with clinical practice, except that we did not include isolates with intermediate minimum inhibitory concentrations among those with specific resistance pattern combinations (eg, resistance to ampicillin, streptomycin, sulfonamide, and tetracycline). CLSI interpretive criteria for ciprofloxacin changed in 2012 to reflect clinical significance; susceptibility was defined as ≤0.06 µg/mL, intermediate resistance as 0.12–0.5 µg/mL, and resistance as ≥1.0 µg/mL.

We analyzed data on Salmonella isolated from 2003 through 2013. Patients with Salmonella isolated from blood and stool were compared by sex and age, using odds ratios (ORs) and 95% confidence intervals (CIs). Patients were divided into 6 age groups: <1, 1–4, 5–17, 18–64, 65–84, and ≥85 years. Non-Typhimurium blood and stool isolates were compared with serotype Typhimurium, which served as the referent group. Serotypes with <10 blood isolates were combined into a category called other. Blood and stool isolates were compared for antimicrobial resistance. We determined the number of isolates that were fully susceptible, resistant to ≥1 agent, resistant to ≥3 CLSI classes, and resistant to ≥5 CLSI classes. We also examined the following resistance pattern combinations: resistance to ampicillin, streptomycin, sulfonamide, and tetracycline but not chloramphenicol; resistance to ampicillin, chloramphenicol, streptomycin, sulfonamide, and tetracycline; resistance to ampicillin, chloramphenicol, streptomycin, sulfonamide, tetracycline, amoxicillin–clavulanic acid, and ceftriaxone; and resistance to ampicillin, amoxicillin–clavulanic acid, and ceftriaxone. Resistance to a first-line treatment agent was defined as resistance to ≥1 of the following agents used to treat Salmonella infections: ampicillin, ceftriaxone, ciprofloxacin, and TMP-SMX. Salmonella isolates from blood were compared with stool isolates for resistance to ≥1 agents by serotype using logistic regression.

Various regression models, including robust and logistic, were used to assess the sensitivity of the model choice and to identify specific trends. Estimated annual trends with associated CIs were then computed by agent and specimen source.

RESULTS

During 2003–2013, we tested 23 761 Salmonella isolates. Of those, 21 390 were from blood or stool, 1388 (5.9%) were from urine, and the remaining 983 (4%) were from abscesses, gallbladders, wounds, or other sites. Overall, 524 blood and stool isolates were excluded from analysis; 264 were not serotyped, 200 were partially serotyped, and 60 were rough or nonmotile. Of the 20 866 remaining isolates, 1189 (5.7%) were from blood, and 19 677 (94.3%) were from stool. Of 19 362 isolates with information on sex, 9774 (50.5%) were from female patients. Among 1117 blood isolates for which information on patient sex was available, 608 (54.4%) were from male patients, who were more likely to have bacteremia (OR, 1.23; 95% CI, 1.09–1.39).

The median age was 43 years (range, <1 to 98 years) among patients with blood isolates, compared with 22 years (<1 to 103 years) among those with stool isolates (P < .001). Persons aged 65–84 years (OR, 2.04; 95% CI, 1.73–2.41) or ≥85 years (OR, 2.10; 95% CI, 1.47–3.00) were more likely to have bacteremia than those in the 18–64-year age group (referent). Infants <1 year old (OR, 0.77; 95% CI, .62–.97) and children aged 1–4 (OR, 0.63; 95% CI, .52–.76) or 5–17 years (OR, 0.68; 95% CI, .56–.82) were less likely to have bacteremia.

The proportion of Salmonella isolates that came from blood varied by serotype, ranging from 87.0% for serotype Dublin to 2.3% for serotype Muenchen (Table 1). Compared with Typhimurium, the 3 serotypes most highly associated with bacteremia were Dublin (OR, 128.2; 95% CI, 57.19–287.40), Sandiego (OR, 4.90; 95% CI, 2.90–8.29), and Schwarzengrund (OR, 3.44; 95% CI, 2.14–5.55). Serotypes Enteritidis, Heidelberg, Oranienburg, Panama, Poona, and Rubislaw were also more likely to be isolated from blood, and serotypes Javiana, Muenchen, and Newport were less likely to be isolated from blood.

Blood isolates were further characterized by resistance pattern; 868 (73.0%) of 1189 blood isolates were susceptible to all agents tested. Among the 321 blood isolates that were resistant to ≥1 agent, resistance to a first-line treatment agent was found in 237 (73.8%). Compared with stool isolates, blood isolates were associated with resistance to ≥1 agent, ≥3 classes, ≥5 classes, and first-line treatment agents. Isolation from blood was also significantly associated with resistance to most other agents and resistance pattern combinations (Table 2).

Of the 237 blood isolates with resistance to a first-line treatment agent, the most common serotypes were Typhimurium (85 isolates), Heidelberg (38 isolates), and Dublin (33 isolates). Serotypes Dublin, Heidelberg, and Newport had the greatest number of blood isolates resistant to ceftriaxone. Serotypes Typhimurium, Heidelberg, and Dublin had the greatest number resistant to ampicillin, serotype Enteritidis had the greatest number resistant to ciprofloxacin, and serotype Typhimurium had the greatest number resistant to TMP-SMX (see Supplementary Data).

The following serotypes had the strongest association between resistance to ≥1 agent and isolation from blood: Panama (OR, 6.85; 95% CI, 1.25–37.58), Javiana (OR, 3.76; 95% CI, 1.08–13.04), Typhimurium (OR, 2.04; 95% CI, 1.52–2.73), and Enteritidis (OR, 1.61; 95% CI, 1.13–2.29) (Table 3). In contrast, for serotype I 4,[5],12:i:−, stool isolates were significantly more likely than blood isolates to be associated with resistance to ≥1 antimicrobial agent.

During 2003–2013, ciprofloxacin resistance increased and ampicillin resistance decreased for both blood and stool isolates, though the trends were statistically significant only for stool isolates (Table 4). In contrast, ceftriaxone resistance increased significantly for blood isolates (OR, 1.12; 95% CI, 1.02–1.22) and decreased significantly for stool isolates (OR, 0.95; 95% CI, .92–.97). Resistance to TMP-SMX did not change significantly for blood or stool isolates.

DISCUSSION

We found a substantial amount of resistance to antimicrobials used for Salmonella bloodstream infections. Overall, approximately 5% of all Salmonella isolates that came from blood were resistant to ceftriaxone. Resistance to ceftriaxone has doubled since a 1996–2007 study from NARMS. In that study, only 2.5% of all Salmonella isolates that came from blood were resistant to ceftriaxone [12]. Ceftriaxone is considered a first-line treatment for Salmonella bacteremia, and increasing antimicrobial resistance is concerning for clinical treatment and patient outcomes. If this path of increasing resistance continues, we may soon be at a crossroads where first-line treatment recommendations will need to change, as occurred with gonorrhea treatment and the use of cefixime; resistance increased from 0.1% to 1.5% over 5 years and prompted a change in gonorrhea treatment guidelines [22].

Fluoroquinolones, penicillins, and cephalosporins are commonly prescribed for a variety of clinical syndromes, and increasing human exposure to these antimicrobials may lead to an increased risk of antimicrobial resistance [10, 15, 23]. Since the previous NARMS study, fluoroquinolone resistance has nearly doubled. We also found that resistance to ampicillin, amoxicillin–clavulanic acid, or cefoxitin was associated with Salmonella bacteremia. Nalidixic acid resistance correlates with resistance to ciprofloxacin and may predict treatment failure [12, 14]. We found that 4% of all Salmonella isolates that came from blood had nalidixic acid resistance and 4.5% had ciprofloxacin resistance; the NARMS study from 1996–2007 found that 2.7% of all Salmonella isolates that came from blood were resistant to nalidixic acid [12]. Resistance to ≥1 agent, and resistance to ≥3 or ≥5 classes of antimicrobials were also associated with bacteremia, supporting the finding that bacteremia is more common in drug-resistant infections than susceptible ones [10–12]. We do not know the relative contribution of each biological or clinical mechanism that may link antimicrobial resistance to bloodstream infections; these infections might be due to the failure or reduced efficacy of empirical antimicrobial treatment which would result in more severe illness, the presence of additional virulence factors that could enhance invasiveness and worsen patient outcome, or the fact that patients whose isolates are submitted to NARMS for testing are inherently more likely to be hospitalized and seek care from a provider owing to increased severity of symptoms [11].

The annual proportion of isolates that came from blood with ceftriaxone resistance increased from 2003 through 2013, whereas the proportion of stool isolates with ceftriaxone resistance decreased. This finding did not hold for the other first-line treatment agents. The reason that blood isolates are increasingly resistant to ceftriaxone whereas stool isolates are becoming more susceptible to ceftriaxone is probably related to serotype. It is likely that the overall distribution of Salmonella serotypes in blood and stool is constantly evolving, and serotypes that are commonly resistant may be increasing disproportionally in blood over time. For example, host adapted, highly resistant serotypes have become associated with Salmonella blood infections, particularly serotypes Dublin and Choleraesuis [24–26]. Salmonella Choleraesuis, a serotype that is host adapted to swine, is becoming increasingly resistant, probably owing to various resistance genes and plasmids [25, 26], possibly acquired through agricultural antimicrobial use.

In our study, bacteremia was most common among patients with serotypes Dublin, Sandiego, Schwarzengrund, Poona, Panama, Heidelberg, Oranienburg, Rubislaw, and Enteritidis, compared with serotype Typhimurium. These serotype-specific findings are consistent with previous studies, with the notable exceptions of serotypes Poona and Rubislaw. A previous study showed a higher risk of invasive disease for serotype Poona compared with Typhimurium [6], but other studies have not supported this association [11, 12, 27]. Previous studies have not shown an association between bacteremia and serotype Rubislaw [6, 11, 12, 27]. The reason for the incongruous findings for serotype Poona and the new association of bacteremia with serotype Rubislaw is unknown. Most of the infections with serotypes Rubislaw and Poona in our study were among children. Both these serotypes have an historical association with reptiles [28], and reptile-associated Salmonella infections may place children at an increased risk of invasive infection [29]. This may explain the increased blood isolation of Rubislaw and Poona isolates in our study and may indicate that reptile exposures predispose to Salmonella bacteremia with certain serotypes.

We confirmed findings from previous studies regarding risk factors for Salmonella bacteremia, including a higher frequency of Salmonella bacteremia in men than in women, a higher median age among patients with bacteremia than among patients with stool isolates, and the highest rate of bacteremia among persons aged ≥64 years, consistent with invasive disease being more common in older adults [12]. Other risk factors, such as international travel, have been associated with antimicrobial resistance in patients with Salmonella bacteremia, [30], but travel data are not collected in NARMS. This study and a previous one that used NARMS data [12] found that the odds of Salmonella bacteremia was lower among infants aged <1 year than among adults aged 18–64 years. This may be due in part to differences in Salmonella serotype distribution. For example, in our study a lower proportion of infants than adults had Salmonella Enteriditis, which is known to be an invasive serotype [6, 27]. Moreover, in contrast to adults with Salmonella gastroenteritis, it is recommended that infants <3 months of age receive treatment to avert invasive disease [31]. Further investigation is needed to characterize the vulnerability of infants to Salmonella bacteremia.

This study had some limitations. The NARMS protocol for the state and local laboratories requires submission of every 20th Salmonella isolate to the CDC. It is possible that some patients who had a stool isolate submitted to CDC NARMS also had bacteremia, but the stool isolate was the 20th isolate submitted by the health department to CDC NARMS. This would bias our associations toward the null. Differential resistance among the missing and interaction between variables, possibly regarding both serotype and age, may influence results. We were unable to examine serotypes with regard to virulence as an explanation for their propensity cause bacteremia (or lack thereof) to. NARMS does not capture patient treatment or outcome data. For serotypes with small numbers of isolates, limited precise estimates of antimicrobial resistance could not be made because of limited power. We were also limited by the short time series of 11 years (2003–2013) to perform time trends analysis.

In conclusion, Salmonella blood isolates were more likely than stool isolates to be resistant to ≥1 agent and to first-line treatment agents. Resistance to first-line treatment agents in patients with Salmonella bacteremia is a concern for public health and clinical outcomes and is important for informing clinical decisions regarding appropriate treatment. Judicious antimicrobial use in both humans and food-producing animals is crucial to limit the emergence and spread of resistance.

Notes

Acknowledgments. We acknowledge the collaborative effort among the Centers for Disease Control and Prevention, the Food and Drug Administration (FDA), and the US Department of Agriculture in conducting National Antimicrobial Resistance Monitoring System (NARMS) surveillance. We appreciate the efforts of state and local health departments that contribute regularly to NARMS and thank them for their participation.

Financial support. NARMS is funded in part by the US FDA.

Potential conflicts of interest. All authors: No potential conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References

Author notes