Extraintestinal focal infections in adults with nontyphoid Salmonella bacteraemia: predisposing factors and clinical outcome


Wen-Chien Ko MD, Department of Internal Medicine, National Cheng Kung University Hospital, No. 138, Sheng Li Road, Tainan 704, Taiwan.
(fax: 886 6 2752038; e-mail: winston@mail.ncku.edu.tw).


Background.  Nontyphoid Salmonella (NTS) isolates lead to not only self-limited, acute gastrointestinal infections, but also bacteraemia with or without extraintestinal focal infections (EFIs). The risk factors associated with EFIs in adults with NTS bacteraemia were not clearly elucidated.

Methods.  In a medical center in southern Taiwan, patients aged ≥18 years with NTS bacteraemia between January 1999 and June 2005 were included for analysis.

Results.  Of 129 patients, 51 (39.5%) were complicated with EFIs. The most common EFI was mycotic aneurysm, followed by pleuropulmonary infections and spinal osteomyelitis. Compared to patients with primary bacteraemia, those with EFIs had higher leucocyte counts (P = 0.004) and higher serum levels of C-reactive protein (P < 0.0001). The development of EFIs was associated with a higher mortality, more severe septic manifestations, longer hospital stays and duration of antimicrobial therapy. Univariate analysis revealed that diabetes mellitus (P = 0.02), hypertension (P = 0.02) and chronic lung disease (P = 0.006) were significantly associated with EFIs. However, patients with malignancy (P = 0.01) and immunosuppressive therapy (P = 0.03) were less likely to develop EFIs. On the basis of multivariate analysis, an independent factor for the occurrence of EFIs was age [adjusted odds ratio (aOR) 1.05; 95% confidence interval (CI) 1.02–1.07; P < 0.0001], whilst malignancy was negatively associated with EFIs (aOR 0.16; 95% CI 0.14–0.78; P = 0.01).

Conclusion.  Amongst patients with NTS bacteraemia, EFIs often occurred in the aged, and were associated with a higher mortality and morbidity. Recognition of specific host factors is essential for identification of EFIs which often demand early surgical interventions and prolonged antimicrobial therapy.


Nontyphoid Salmonella (NTS) isolates are commonly associated with food contaminants and cause human infections [1, 2]. Although most patients with NTS infection have self-limited gastroenteritis, about 5% develop subsequent bacteraemia [1]. Focal complications of NTS bacteraemia are identified in 8–16.7% of affected patients [3–5]. In adults, NTS bacteraemia poses significant health threats because affected patients usually have underlying diseases [5–7]. Extraintestinal focal infections (EFIs) included septic aortitis, meningitis, pneumonia, septic arthritis, osteomyelitis or cholangitis [1, 8]. Such complications often required prolonged antimicrobial therapy and surgical intervention.

The prognosis for children with NTS bacteraemia is excellent [4, 6], but the reported mortality for adults with NTS bacteraemia is higher [3, 6, 9, 10]. Predictors for mortality included advanced age and extraintestinal organ involvement [3, 6]. Amongst EFIs caused by NTS, endovascular infection posed high mortality and morbidity [8, 11]. The risk factors for vascular infection in adult patients with NTS bacteraemia described previously included aged ≥50 years, underlying atherosclerosis and infections caused by S. Choleraesuis [11–13]. The mortality rate of focal infections caused by NTS bacteraemia was reported as 50% amongst patients with acquired immunodeficiency syndrome (AIDS) [14]. However, most of the studies about EFIs caused by NTS are either reports of cases or involved paediatric patients [3–5, 14]. Notably, certain serotypes of NTS, i.e. Choleraesuis, Dublin and Virchow, were more likely to cause bacteraemia and develop EFIs in different endemic areas [15–17]. Meanwhile, antimicrobial resistance to broad-spectrum cephalosporins or fluoroquinolones in S. Choleraesuis has been reported in recent years and become therapeutic challenge [18–20].

Nontyphoid Salmonella bacteraemia can cause EFIs and is a life-threatening infectious disease in susceptible hosts. Early identification of EFIs in patients with specific risk factors will prompt adequate surgical and medical treatment. This study was undertaken to evaluate patients with NTS bacteraemia and to delineate the risk factors for EFIs.

Materials and methods

A retrospective study was conducted in a university hospital, which is a medical centre with 900 beds, and serves a population of about two million inhabitants in southern Taiwan. Patients aged 18 or older, with at least one blood sample positive for NTS identified by the clinical microbiology laboratory between January 1999 and June 2005 were included. Medical records for those patients were reviewed regarding demographic data, underlying diseases, clinical manifestations, laboratory data on admission, antimicrobial treatment, clinical outcome and the length of hospital stay. For patients with recurrent NTS bacteraemia, only the first episode was included as the indicated infection.

Microbiology and antimicrobial susceptibility

The serogroup of Salmonella isolates was determined by O antisera (Difco Laboratories, Detroit, MI, USA) first by the slide agglutination test. For serogroup C or D isolates, the presence of Vi antigen in S. Typhi or S. Paratyphi was screened by Vi antisera (BBL, Cockeysville, MD, USA). Salmonella Choleraesuis was identified if the citrate test was negative in the isolates of serogroup C1 [21].

Antimicrobial susceptibilities of trimethoprim-sulfamethoxazole (TMP-SMX), ampicillin and chloramphenicol were determined by the disc-diffusion method, and those of nalidixic acid, ciprofloxacin, ceftriaxone, cefepime and meropenem by Etest strips (AB Biodisk, Solan, Sweden). The interpretations of susceptibility data followed the criteria proposed by the Clinical and Laboratory Standards Institute [22].


The infectious focus of bacteraemia was determined clinically by the presence of an active infection site coincident with NTS bacteraemia, or the isolation of an identical organism from clinical specimens other than blood and faeces. Patients were considered to have primary bacteraemia if there was no clinical evidence of other infectious focus and no identical organism isolated from a clinical specimen other than blood and faeces.

Fever was defined as an axillary body temperature ≥38 °C. The severity of illness was evaluated on the first day of onset of bacteraemia based on Pittsburgh's bacteraemic score [23]. Patients were assumed to have diabetes mellitus if the fasting serum glucose level was >126 mg dL−1. Renal insufficiency was defined as a serum creatinine ≥1.5 mg dL−1. Liver cirrhosis was defined by abdominal ultrasonography and clinical follow-up findings. Immunosuppressive therapy was defined as the receipt of corticosteroid, i.e. 10 mg prednisolone per day or equivalent dosage for at least 2 weeks, chemotherapy for malignancy or immunosuppressive agents for organ transplantation within 1 month prior to admission.

Shock was defined as a systolic blood pressure ≤90 mmHg or an unstable haemodynamic status requiring inotropic agents to maintain blood pressure. Severe sepsis was defined as sepsis with organ dysfunction on the basis of Sepsis Definitions Conference, 2001 [24]. Antimicrobial therapy was considered to be appropriate, if the aetiological pathogen was susceptible in vitro to at least one of the drugs administered within 3 days after the onset of bacteraemia. An episode of bacteraemia was considered to be ‘persistent’ when blood cultures yielded the identical organism in patients with appropriate antimicrobial therapy for at least 72 h. The length of hospital stay and duration of antibiotics usage were evaluated in survivors.

Statistical analysis

To identify factors associated with extraintestinal infections, patients with extraintestinal infections were compared with those with primary bacteraemia. Data analysis was conducted using the Statistical Package for the Social Science for Windows (SPSSWIN; SPSS, Chicago, IL, USA), Version 12.0. Continuous variables were expressed as means ± standard deviation (SD) and were compared by the Student's t, Mann–Whitney or one-way anova tests. Categorical variables, expressed as numbers and percentages, were compared using the chi-squared test or Fisher's exact test. Multivariate analysis was performed with the stepwise logistic regression model, on the variables with P≤0.1 as the limit for entering or removing terms. Survival rate was calculated by the Kaplan–Meier method, and the significance was evaluated by the log-rank test.


Clinical information

During the 78-month study period, 145 episodes of NTS bacteraemia occurred in 130 adults (84 men and 46 women), and accounted for an average incidence of 7.25 cases per 100 000 adult discharges during the study period. One patient was excluded for analysis due to incomplete medical information. The distribution of serogroups of NTS isolates was in the order of groups B (56 isolates, 43.4%), C (45, 34.9%), D (26, 20.2%) and E (2, 1.6%). Of serogroup C Salmonella isolates, 41 (91.1%) serogroup C1 isolates were serotype Choleraesuis, which accounted for 31.8% of 129 bacteraemic Salmonella isolates. Clinical characteristics of 129 adults with NTS bacteraemia are shown in Table 1. Their mean age was 56.5 ± 18.2 years (range 19–98) with a male preponderance.

Table 1.   Clinical characteristics of 129 adults with nontyphoid Salmonella bacteraemia
CharacteristicsCases, n (%)
  1. Note: A case may have more than one underlying diseases or extraintestinal focal infections. aAll 41 serogroup C1 Salmonella isolates were serotype Choleraesuis. bInclude infective endocarditis (one case), Port-A catheter-related infection (one) and soft-tissue infection (one).

Age, years56.5 ± 18.2
Gender, male83 (64.3)
 B56 (43.3)
 C/C1a45/41 (34.9/31.8)
 D26 (20.2)
 E2 (1.6)
Underlying diseases125 (96.9)
 Malignancy47 (36.4)
 Diabetes mellitus38 (29.5)
 Immunosuppressive therapy37 (28.7)
 Renal insufficiency37 (28.7)
 Hypertension36 (27.9)
 Human immunodeficiency virus infection9/48 (18.6)
 Connective tissue disorders20 (15.5)
 Liver cirrhosis18 (14.0)
 Coronary heart disease13 (10.1)
 Congestive heart failure12 (9.3)
 Chronic lung disease7 (5.4)
Extraintestinal focal infections51 (39.5)
 Mycotic aneurysm19 (14.7)
 Pneumonia/empyema13 (10.1)
 Spinal osteomyelitis7 (5.4)
 Spontaneous bacterial peritonitis3 (2.3)
 Liver abscess2 (1.6)
 Splenic abscess2 (1.6)
 Septic arthritis2 (1.6)
 Othersb3 (2.3)

Of 129 patients, 125 (96.9%) had at least an underlying medical illness. Amongst 20 patients with connective-tissue disorders, 16 had systemic lupus erythematosus (SLE); the other four had dermatomyositis, Sjogren syndrome, rheumatoid arthritis and psoriasis, respectively, and 17 of whom received immunosuppressive therapy for their underlying connective-tissue disorders. Of 48 patients tested for human immunodeficiency virus (HIV) infections, nine (18.6%) were infected by HIV-1, and their median CD4 count at initial presentations was 16 cells mm−3 (range 0–36), therefore regarded as cases of AIDS.

Extraintestinal focal infections

Fifty-one (39.5%) patients had EFIs, including mycotic aneurysm, pneumonia, pulmonary empyema, osteomyelitis, hepatic or splenic abscess, septic arthritis, spontaneous bacterial peritonitis, catheter-related infection and infective endocarditis (Table 1). At 90 days after the onset of Salmonella bacteraemia, those with EFIs had a worse prognosis than those with primary bacteraemia (Fig. 1). Mycotic aneurysm was the most common EFI, followed by pulmonary/pleural infection and osteomyelitis. Nineteen patients (37.3%) developed mycotic aneurysm (11 with thoracic aorta involvement, seven with abdominal aorta, one with iliac artery), and 13 patients were fatal. Salmonella Choleraesuis accounted for 52.6% (10 cases) of endovascular infection. Ten of them underwent surgical interventions, and five died. In contrast, eight of nine patients not receiving surgery died. The only one survivor was an 84-year-old male patient with infected thoracic aortic aneurysm who received antimicrobial therapy for an year, and was still alive 3 months later after discontinuing treatment.

Figure 1.

 Kaplan–Meier survival analysis of cases of nontyphoid Salmonella bacteraemia with and without extraintestinal focal infections (the latter were referred to be ‘primary bacteraemia’) until 90 days after the onset of bacteraemia.

A total of 13 patients had NTS pulmonary or pleural infections, and four died. Three of them were pulmonary empyema according to positive effusion cultures, and the others were diagnosed as bacteraemic pneumonia on the basis of blood cultures and pulmonary infiltration on radiographic image studies. Seven patients had NTS spinal osteomyelitis, and six of them involved S. Choleraesuis infection. The most common infection site was lumbar spine (four cases), followed by thoracic spine (two cases) and cervical spine (one case). Five of them were complicated with paraspinal abscess, and three with epidural abscess. Six patients received surgical interventions, and one died 45 days after surgical and medical therapy.

Amongst 78 patients with primary NTS bacteraemia, 35 (44.9%) had underlying malignancy, and 17 (21.8%) had connective-tissue disorders. Eight of nine HIV-infected patients had primary bacteraemia, and finally recovered with antimicrobial treatment.

Risk factors for EFIs

The clinical characteristics and laboratory data of patients with EFIs and those with primary bacteraemia are compared in Tables 2 and 3. EFIs occurred more frequently in patients with increasing age, diabetes mellitus, hypertension and chronic lung disease. However, EFIs were less developed in patients with malignant diseases, connective-tissue disorders and immunosuppressive therapy (Table 2). Those with EFIs more often manifested with shock or severe sepsis, and had higher levels of white blood cell counts and C-reactive protein at admission (Table 3).

Table 2.   Comparisons of clinical categorical variables in cases of nontyphoid Salmonella bacteraemia with or without extraintestinal focal infections
VariablesExtraintestinal focal infectionsP-values
Yes (n = 51)No (n = 78)
  1. Note: Data are presented as the case numbers and their proportions in parentheses, unless indicated specifically. aOnly 48 patients were tested for human immunodeficiency virus infection.

Gender, male36 (70.6)48 (61.5)0.292
Underlying diseases48 (94.1)77 (98.7)1.000
 Diabetes mellitus21 (41.2)17 (21.8)0.018
 Hypertension20 (39.2)16 (20.5)0.021
 Renal insufficiency16 (31.4)22 (28.2)0.700
 Malignancy12 (23.5)35 (44.9)0.014
 Immunosuppressive therapy9 (17.6)28 (35.9)0.025
 Chronic lung diseases7 (13.7)1 (1.3)0.006
 Liver cirrhosis5 (9.8)13 (16.7)0.271
 Connective tissue disorder4 (7.8)17 (21.8)0.050
 HIV infectiona1/19 (5.3)8/29 (27.6)0.068
Clinical signs
 Fever35 (68.6)62 (79.5)0.173
 Tachycardia32 (62.7)45 (57.7)0.531
 Altered consciousness7 (13.7)12 (15.4)0.772
 Shock25 (49.0)16 (20.5)0.001
 Severe sepsis39 (76.5)39 (50.0)0.003
Pittsburgh's bacteraemic score ≥47 (13.5)3 (3.9)0.089
Persistent bacteraemia8 (15.7)6 (7.7)0.154
Appropriate antimicrobial therapy within 72 h43 (84.3)51 (67.1)0.003
Mortality at 90 days after the onset of bacteraemia23 (45.1)19 (24.4)0.014
Table 3.   Comparisons of clinical characters in cases of nontyphoid Salmonella bacteraemia with or without extraintestinal focal infections
Clinical charactersExtraintestinal focal infectionsP-values
Yes (n = 51)No (n = 78)
  1. Values are expressed as mean ± SD or median (interquartile range).

Age, years63.8 ± 16.051.2 ± 17.7<0.0001
C-reactive protein (mg L−1)168.3 ± 138.583.5 ± 88.3<0.0001
Leucocyte (cells mm−3)13829.4 ± 1169136.4 ± 6401.90.004
Haemoglobin (g dL−1)11.4 ± 2.610.7 ± 2.40.129
Platelet ( ×1000 cells mm−3)192.5 ± 111.0190.6 ± 130.80.928
Serum creatinine (mg dL−1)1.62 ± 1.551.86 ± 2.160.495
Aspartate aminotransferase (U L−1)99.9 ± 221.2132.1 ± 295.90.515
Alanine aminotransferase (U L−1)78.9 ± 119.873.7 ± 129.40.821
Glucose (mg dL−1)205.1 ± 199.1169.8 ± 134.70.379
Duration (days)
 Clinical symptoms before admission2.0 (1.0–7.0)2.0 (1.0–7.0)0.436
 Antimicrobial therapy20 (9–42.5)14 (10–21)0.045
 Hospitalization22 (10–49)11 (7–22)0.001

Clinical outcomes of EFIs

Patients with EFIs tended to have more severe illness at initial presentations than those with primary bacteraemia according to Pittsburgh bacteraemic score, although the difference was not statistically significant (Table 2). Furthermore, the former needed longer durations of antimicrobial therapy and hospitalization. Patients with EFIs had a significantly lower survival rate after 90-day follow-up, as revealed by the Kaplan–Meier survival analysis (Fig. 1). Such an unfavourable outcome became evident since the first week after the onset of bacteraemia (34.2% vs. 9.0%, P = 0.01). To study the risk factors for EFIs, several host variables, including age, the presence of diabetes mellitus, hypertension, chronic lung diseases, malignant diseases, or connective tissue disorders, and the receipt of immunosuppressive therapy were included in the multivariate analysis. According to the final model, the risk of EFIs in patients with NTS bacteraemia increased significantly with age, and in contrast, the presence of malignancy was negatively associated with the development of focal infections (Table 4).

Table 4.   Multivariate analysis of host factors for extraintestinal focal infections in cases of nontyphoid Salmonella bacteraemia
VariablesAdjusted odds ratio95% confidence intervalP-values
Connective tissue disorders0.1390.018–1.4390.160
Chronic lung diseases0.1530.017–1.3990.096
Diabetes mellitus0.5650.238–1.3450.197
Immunosuppressive therapy1.0510.340–3.2500.932

Variations of EFIs amongst three major Salmonella serogroups

Certain underlying diseases and clinical manifestations appeared to be associated with a specific serogroup (Table 5). Of note, patients with S. Choleraesuis bacteraemia were rarely found to have underlying malignancy. Although the likelihood of developing EFIs was similar amongst patients with Salmonella bacteraemia caused by the three major serogroups, serogroup C1 bacteraemia was more often complicated with spinal osteomyelitis than serogroup B or D Salmonella bacteraemia (Table 5). Although mycotic aneurysm tended to be associated with S. Choleraesuis infections, the difference is not statistically significant (P = 0.059).

Table 5.   Clinical features of patients with bacteraemia caused by three major serogroups of nontyphoid Salmonella
 Serogroup C1a (n = 41)Serogroup B (n = 56)Serogroup D (n = 26)P-values
  1. aAll 41 serogroup C1 Salmonella isolates were serotype Choleraesuis. bAntimicrobial susceptibilities were determined by disc-diffusion method.

Age, years57.3 ± 20.956.4 ± 15.453.2 ± 19.30.65
Gender, male28 (68.3)34 (60.7)19 (73.1)0.50
Underlying diseases
 Diabetes mellitus15 (36.6)13 (23.2)9 (34.6)0.31
 Hypertension10 (24.4)18 (32.1)5 (19.2)0.43
 Immunosuppressive therapy9 (22.0)21 (37.5)6 (23.1)0.19
 Connective tissue disorders8 (19.5)11 (19.6)2 (7.7)0.36
 Malignancy7 (17.1)26 (46.4)12 (46.2)0.006
 Liver cirrhosis4 (9.8)12 (21.4)2 (7.7)0.15
 Chronic lung diseases5 (12.2)2 (3.6)1 (3.8)0.19
Extraintestinal focal infections20 (48.8)20 (35.7)9 (34.6)0.36
 Mycotic aneurysm10 (24.4)4 (7.1)4 (15.4)0.059
 Osteomyelitis6 (14.6)1 (1.8)0 (0)0.01
 Pneumonia/empyema4 (9.8)5 (8.9)4 (15.4)0.66
Appropriate antimicrobial therapy within 72 h32 (78.0)37 (66.1)19 (73.1)0.39
Mortality at 90 days after bacteraemia onset13 (37.1)19 (33.9)8 (30.8)0.96
Antimicrobial susceptibility
 Chloramphenicol9 (22.0)54 (96.4)23 (88.5)<0.0001
 Ampicillin4 (9.8)16 (28.6)25 (96.2)<0.0001
 Trimethoprim-sulfamethoxazole2 (4.9)18 (32.1)25 (96.2)<0.0001
 Susceptibility to either of three above agentsb11 (26.8)56 (100)26 (100)<0.0001

Antimicrobial susceptibility and therapy in three major Salmonella serogroups

Compared with serogroup B or D Salmonella isolates, lower susceptible rates to ampicillin, chloramphenicol and TMP-SMX were noted amongst serogroup C1 isolates, which was identified to be S. enterica serotype Choleraesuis (Table 5). All serogroup C1 isolates were resistant to nalidixic acid (MIC90 > 256 μg mL−1), and near a half (46.3%) of these isolates were nonsusceptible to ciprofloxacin (Table 6). Resistance to ceftriaxone was present in <5% of serogroup C1 isolates, but absent in serogroup B or D isolates. Cefepime and meropenem were active against all isolates.

Table 6. In vitro antimicrobial susceptibility of 123 nontyphoid Salmonella (NTS) isolates based on different serogroups
Antimicrobial agentsMIC50/MICinline image(μg mL−1), susceptible isolate no. (%)
Serogroup C1b (n = 41)Serogroup B (n = 56)Serogroup D (n = 26)P-values
  1. aMIC50 and MIC90 represent the minimal drug concentration inhibiting 50% and 90% of isolates tested. bAll serogroup C1 NTS isolates were serotype Choleraesuis.

Nalidixic acid≧256/≧256, 0 (0)4/≧256, 45 (80.4)4/≧256, 21 (80.8)<0.001
Ciprofloxacin8/≧32, 22 (53.7)0.016/0.25, 53 (94.6)0.016/0.125, 26 (100)<0.001
Ceftriaxone0.064/2, 39 (95.1)0.064/0.094, 56 (100)0.064/0.094, 26 (100)0.417
Cefepime0.125/0.38, 41 (100)0.125/0.25, 56 (100)0.094/0.125, 26 (100)
Meropenem0.023/0.032, 41 (100)0.032/0.064, 56 (100)0.023/0.032, 26 (100)

With regard to antimicrobial treatment, the majority (71.5%, 88/123) of three major serogroups had been empirically treated by at least one appropriate drug, often a third-generation cephalosporin (ceftriaxone or cefotaxime, 56.8%) or a fluoroquinolone (ciprofloxacin or levofloxacin, 19.3%), within 72 h. Amongst patients with bacteraemia caused by three major Salmonella serogroups, the appropriateness of variety of empirical antimicrobial therapy did not differ greatly. Moreover, the appropriateness of empiric antimicrobial therapy could not warrant the decrement in development of EFIs, as suggested by that appropriate antimicrobial agents have been empirically given to 84% of 51 patients with EFIs, and only 65% of 78 patients without EFIs were empirically treated by appropriate drugs (P = 0.003, Table 2).


The elderly with age-related disorders were at the risk of developing EFIs in our study. Moreover, the most common EFI was mycotic aneurysm, which accounted for 36.5% of total EFIs, and for 22.4% amongst aged patients (≥50 years old). The risk factors for vascular infection in adult patients with NTS bacteraemia described previously included aged ≥50 years, underlying atherosclerosis and infections caused by S. Choleraesuis [11–13]. A Taiwanese study revealed that the incidence of mycotic aneurysm in aged patients with NTS bacteraemia was as high as 40.6% [11], which is higher than that observed in other areas [6, 12, 13]. This phenomenon is reasonably related to the high prevalence rate of human infections caused by S. Choleraesuis, which has been the second common serotype isolated from humans in Taiwan [15, 25]. EFIs other than mycotic aneurysm in our study included pleuropulmonary infections, spinal osteomyelitis and visceral abscesses. These severe complications led to a high mortality and morbidity rate, and often needed surgical intervention besides antimicrobial therapy. In a Spanish study of 172 cases of salmonella bacteraemia, Galofre et al. [3] reported that septic metastasis is associated with rapidly fatal underlying diseases (P = 0.001) and leucocytosis (P = 0.001) in the univariate analysis. Moreover, they found that the most common EFI was pulmonary or pleural involvement (17 cases), followed by bone (four cases) and kidney (three cases) infections. However, most of their patients were infected with S. Enteritidis and Typhimurium, and were young people with AIDS, or recipients of kidney or bone marrow transplant. These host and microbial factors were different from those in the present study, and might be, to some extent, related to the geographical variations of EFIs.

In our study, there were seven cases complicated with osteomyelitis. All were aged and had spinal involvement. Noteworthy, most (six patients) of them had S. Choleraesuis bacteraemia (Table 5). None of our affected patients had the underlying haemoglobinopathy, which has been described as a risk factor for Salmonella osteomyelitis in young people, amongst whom the frequently affected sites were ribs, spine and long bones [26, 27]. Therefore, our study highlighted another clinical pattern of Salmonella-related bone infection, spinal osteomyelitis in the elderly without haemoglobinopathy. However, due to limited cases in our series, more clinical observations are required.

Most patients with NTS bacteraemia have extremes of age and severe immunosuppression, such as malignancies, HIV infection or connective-tissue disorders [8–10]. Our analysis indicated that underlying malignancy was an independent negative predictor of EFIs for patients with NTS bacteraemia. Moreover, patients with EFIs had a lower prevalence of connective-tissue disorders and receipt of immunosuppressive agents on univariate analysis, although the difference did not achieve a statistical significance on the multivariate analysis. In the study of Hsu et al. in Taiwan [11], immunodeficiency and SLE contributed to the development of primary bacteraemia. Interestingly, immunodeficiency and solid-organ malignancy were significant negative predictors for endovascular infection on multivariate analysis. In studies of SLE patients with Salmonella infection, SLE was indeed a substantial risk of bacteraemia in those patients, but focal infections were rarely noted [28, 29]. The reason that local infections occur rarely in these immunodeficient patients with NTS bacteraemia remains undefined. One possible explanation was that those patients had been the subjects of long-term follow-up in the hospital and were advised to seek health care whenever they felt discomfort. In the present study, patients with malignant diseases in average sought medical care 2 days earlier than those without malignant diseases (P = 0.021).

Of more than 2000 nontyphoid Salmonella serotypes, S. Choleraesuis was believed to exhibit the highest level in invasiveness (determined in terms of invasiveness index, i.e. number of extraintestinal infections divided by total number of infections), and is usually associated with bacteraemia and invasive extraintestinal infections in humans [15, 16]. Nevertheless, we found that patients with S. Choleraesuis bacteraemia were rarely found to have underlying malignancy. As the interaction between the immune system of immunocompromised hosts and some invasive serotypes of NTS is not clear, further studies are needed.

The EFIs of NTS bacteraemia will reasonably cause not only significant mortality and morbidity, but also result in a longer duration of antimicrobial therapy and hospitalization. Patients presenting with leucocytosis and elevated C-reactive protein levels warrant detailed examinations to rule out any possible EFIs. Although most patients with EFIs (43/51, 84.3%) had been treated with in vitro active antimicrobial therapy within 72 h of admission, their clinical outcome did not improve accordingly. Most of our patients received antimicrobial treatment with a third-generation cephalosporin because the prevalent S. Choleraesuis expressed high-level resistance to conventional antimicrobial agents and ciprofloxacin in Taiwan [21, 30, 31]. EFIs had been linked to a poor prognosis, which was related to the fact that the affected patients were older and often had severe systemic complications (severe sepsis or shock) [11].

Salmonella Choleraesuis isolates in the study were more resistant to certain antimicrobial agents than serogroup B or D isolates, which had been recognized before [21, 30]. Increasing antimicrobial resistance in the particularly virulent Salmonella serotype would limit the list of available alternatives of antimicrobial therapy. In Taiwan, for severe EFIs that may be caused by S. Choleraesuis, fluoroquinolones might not be considered empirically, unless they are proved to be active in vitro.

Our study has some limitations. Because of the retrospective nature, clinical data in the medical records might be incomplete. Although extensive serotyping of all surface antigens can be used for formal identification by the scheme of Kauffman [32], most clinical microbiological laboratories perform a few agglutination reactions to define specific O antigens into serogroups (A, B, C1, C2, D and E). Therefore, identification of Salmonella species was only limited to serogroups, and Vi antisera have been utilized to exclude S. Typhi and Paratyphi in the present study. However, S. Choleraesuis was locally prevalent in Taiwan, and serogroup C1 Salmonella isolates could be identified as serotype Choleraesois if citrate test was negative, which scheme has been validated in earlier surveys in Taiwan [21, 33].

In summary, NTS bacteraemia should be considered as a severe and potentially life-threatening infectious disease in aged patients. Recognition of specific host factors in patients with NTS bacteraemia is helpful for early identification of EFIs, which warrants surgical intervention or prolonged antimicrobial therapy for a favourable outcome.

Financial support


Conflict of interest statement

No conflict of interest was declared.