Bloodstream Infection in Adults With Sickle Cell Disease

Association With Venous Catheters, Staphylococcus aureus, and Bone-Joint Infections

Abstract

Although well documented in children with sickle cell disease (SCD), the incidence, cause, and outcome of bloodstream infection (BSI) are poorly defined in adults with SCD. Through a 5-year retrospective analysis of a cohort of 900 patients followed at our institution, we identified 56 episodes of BSI in 47 patients. The incidence rate of BSI was 1.2 episodes per 100 patient-years. As compared to the patients followed in the cohort, those with BSI were more likely to be younger (p = 0.001), to have Hb-S disease (p = 0.008), severe disease (p = 0.001), or additional immunosuppression (p = 0.05). BSI was hospital-acquired in 46% of cases and mainly associated with venous catheters (41%) and Staphylococcus aureus (34%). Pneumococci were rarely identified (10.7%). Despite an adequate duration of antibiotic therapy, the course of BSI was marked by a high frequency of associated bone-joint infection. Bone-joint infection was noted in 18 patients (32% of episodes) and occurred either during the initial BSI episode (13 patients) or 1-6 months after BSI resolution (5 patients). Factors associated with the occurrence of bone-joint infection were previous osteonecrosis (relative risk, 2.5; 95% confidence interval, 1.2-5.3) and S. aureus infection (relative risk, 3.8; 95% confidence interval, 1.8-8.4). In conclusion, BSI is a rare event in adults with SCD compared to children. It mainly occurs in those with a severe underlying disease and a venous catheter. These patients have a high risk of associated bone-joint infection and therefore must be closely monitored.

Abbreviations: BSI = bloodstream infection, SCD = sickle cell disease, VOC = vaso-occlusive crisis.

INTRODUCTION

Bloodstream infection (BSI) is a frequent and severe complication in young children with homozygous sickle cell disease (SCD), who are highly susceptible to pathogens such as Streptococcus pneumoniae, Haemophilus influenzae, and Salmonella species²⁸. Several mechanisms are believed to contribute to this increased risk, such as asplenia, alteration in complement activity, and abnormalities in the neutrophil responses to cytokines^18,20,26. Life expectancy has increased in patients with SCD: half of them now survive beyond the age of 50 years²⁴. SCD remains a major cause of morbidity in adults, characterized by a high frequency of vaso-occlusive crisis (VOC). Although overt organ failure and acute sickle crisis are the main cause of death in adults with SCD, 1 study showed that bacterial infections represent the second cause of mortality in this population²². However, little is known about BSI in adults in terms of incidence, clinical presentation, risk factors, and complications¹⁵. These patients may be at risk of hospital-acquired BSI, because they often require intravenous therapy and red blood cell transfusion for the management of VOC. Infections can also trigger or mimic a VOC, thus empirical antibiotic treatment is frequently administered in this situation, increasing the risk of selection of resistant microorganisms. As bone infarcts and osteonecrosis are frequent complications of VOC, these patients may also be at risk of associated bone-joint infection during the course of BSI.

In the current study we describe consecutive episodes of BSI in a large cohort of adult patients with SCD followed at a single center. Our specific objectives were to define the incidence, causes, and outcomes of BSI. We also compared patients with and without bone-joint infection to assess factors associated with this complication.

PATIENTS AND METHODS

We retrospectively identified all BSI episodes occurring in the cohort of patients (aged ≥18 yr) followed at the SCD center of our hospital from September 1997 to June 2002 by reviewing the Microbiology Department's computerized database. Approval of the study protocol was obtained from the institutional review board of our hospital.

Patients

According to international definitions of sepsis, all patients had a clinically significant BSI: the same microorganism was isolated from 2 separate blood cultures, or a microorganism was isolated from a single blood culture and either the same microorganism(s) was isolated from a clinically identified source or the patient presented signs or symptoms of systemic inflammatory response syndrome (that is, at least 2 of the following: tachycardia >90/min, tachypnea >20/min or mechanical ventilation, fever >38 °C, hypothermia <36 °C, and a change in blood leukocyte count)². Septic shock was defined as hypotension (<90 mm Hg) or organ dysfunction persisting for at least 1 hour despite administration of fluid, intravascular volume expansion, and/or the use of a vasoactive drug in a patient with BSI.

Blood specimens from which a pathogen was isolated were processed by the Microbiology Department of our institution. Isolates were identified by the use of BACTEC apparatus, and antimicrobial susceptibility testing was performed using antibiotic disks on Mueller Hinton II agar. All tests were done in accordance with the recommendations of the Antibiogram Committee of the French Society of Microbiology⁶.

BSI was considered to be hospital-acquired 1) if it occurred more than 48 hours after admission or 2) if a positive blood culture(s) was drawn from a long-term intravenous device and the patient had been hospitalized within the previous month. All other episodes of BSI were considered to be community-acquired⁸.

Source of BSI

The clinically identified primary source of BSI was recorded on the basis of available clinical and microbiologic information. Digestive, urinary, pulmonary, central nervous system, and soft tissue or bone sources were identified when the same microorganism was isolated from blood culture and a clinically significant source. In the absence of an identified source, BSI was considered to be primary. Cases of documented catheter-related BSI, defined as a catheter blood culture growing the same microorganism(s) as that identified in peripheral blood culture, were considered separately from primary episodes⁴. We distinguished between short-term catheters, which were peripheral or central non-tunneled catheters placed at the time of hospitalization for symptomatic therapy, and long-term catheters, which were tunneled central implantable ports placed before admission and used for long-term transfusion and iron chelation.

Data Collection

For each patient, the following data were recorded from the medical records collected during a routine clinical visit in the year before inclusion: hemoglobin (Hb) genotype (SS, SC, or Sβ° thalassemia), demographic characteristics, long-term therapy modalities of severe disease such as transfusion program, parenteral desferrioxamine or oral hydroxyurea, history of symptomatic osteonecrosis (documented by X-ray or magnetic resonance imaging), other cause of immunosuppression (that is, human immunodeficiency virus [HIV] infection, treatment with corticosteroids or other immunosuppressive agents), and prophylactic measures against infections due to Str. pneumoniae (penicillin prophylaxis or vaccination in the past 5 years).

At admission, we recorded the clinical presentation at diagnosis, including duration of symptoms before admission (fever and/or bone pain), and presence of a venous catheter. Duration of intravenous therapy of VOC before occurrence of BSI was recorded. Bone-joint infection was assessed during the course of each episode of BSI. All patients were followed for at least 6 months after completion of antibiotic therapy.

Statistical Analysis

Categorical variables were compared with the chi-square test or with the Fisher exact test. Quantitative values were compared using the nonparametric Mann-Whitney test. Differences were considered to be significant if p < 0.05. All analyses were performed by using the BMDP Statistical Software.

RESULTS

Incidence of BSI and Characteristics of the Cases at Baseline

During the 5-year study period, 900 patients were followed at the SCD center. Fifty-six episodes of BSI occurred in 47 patients (5.2% of the cohort), giving an incidence of 1.2% BSI episodes per 100 patient-years. Among those patients, 41 patients had 1 episode, 4 had 2 episodes, 1 had 3 episodes, and 1 HIV-infected patient had 4 episodes. During the study period, the total number of admissions of SCD patients followed in the cohort was 2732. Ninety-seven percent of admissions occurred in our hospital. The incidence rate of BSI per 1000 admissions was estimated at 20.5 (95% confidence interval [CI], 15.2-25.8). This rate was higher than the rate of BSI observed in adults without SCD admitted to our hospital during this period (12.7 episodes per 1000 admissions; 95% CI, 12.3-13.1; p < 0.001).

The baseline characteristics of the patients followed in the cohort are summarized in Table 1. Compared with the 853 patients who did not develop BSI, patients with BSI were more likely to be younger and to have Hb-S disease. Patients with BSI had a more severe underlying disease, since they were more often receiving long-term therapy (transfusion program and desferrioxamine) and more likely to have a central venous catheter. The mean number of admissions for VOC during the study period was almost 3 times higher in the BSI group (1.43 admission per 100 patients-years) than in the group of cohort patients without BSI (0.56 admission per 100 patients-years). Additional immunosuppression due to therapy with corticosteroids or methotrexate was also more frequent in the BSI group.

TABLE 1: Baseline Characteristics of 900 Patients With Sickle Cell Disease Followed in the Cohort, Including Those With Bloodstream Infection (BSI)

Clinical Findings

Thirty-five patients (62%) were hospitalized for a VOC before the onset of BSI. Among the 56 BSI episodes, 26 (46%) were hospital-acquired (Table 2). The incidence rate of hospital-acquired BSI was higher in patients with SCD (9.5 episodes per 1000 admissions; 95% CI, 5.9-13.1) than in patients without SCD (5.4 episodes per 1000 admissions; 95% CI, 5.1-5.7; p < 0.001).

TABLE 2: Sources of 56 Episodes of Bloodstream Infection in 47 Adults With Sickle Cell Disease

The main source of BSI was a venous catheter (41% of all episodes), including short-term catheters and implantable ports. Thirty-five patients of the cohort had a long-term central venous catheter; there was device infection in 27 of them (77%). Patients with catheter-related BSI had a longer duration of previous catheterization (mean, 7.1 ± 4.6 d) than those with another source of nosocomial BSI (mean, 2.6 ± 1.5 d) (p = 0.04). Pneumonia and meningitis were rarely observed. Episodes of community-acquired BSI (30 episodes) were mainly of digestive or urinary tract origin, while venous catheters accounted for most of the hospital-acquired BSI episodes (see Table 2).

Bacteriologic Findings

Episodes were mostly monomicrobial (96.4%). Two patients had venous catheter-related polymicrobial BSI. Gram-positive cocci and Gram-negative bacilli were responsible for an equal number of episodes (Table 3). Staphylococcus aureus was the most frequently isolated Gram-positive coccus, causing 34% of the episodes. Only 1 isolate was resistant to methicillin; this isolate was hospital-acquired. Pneumococci were responsible for only 6 episodes in 4 patients. One HIV-infected patient had 3 episodes of pneumococcemia. Three of the patients with pneumococcemia had not received the pneumococcal vaccine, but 2 of them were receiving prophylactic penicillin. Four isolates of pneumococci had reduced susceptibility to penicillin, with MIC values ranging from 0.125 to 0.38 mg/L. These isolates were found only in patients receiving long-term penicillin.

TABLE 3: Distribution of Microorganisms Identified in BSI According to Origin of Acquisition

The relative proportion of Gram-positive cocci and Gram-negative bacilli was similar in community- and hospital-acquired episodes. However, the distribution of species differed markedly within each category. Str. pneumoniae and Streptococcus species were only found in community-acquired cases of BSI. Conversely, S. aureus was the only Gram-positive bacterium found in nosocomial cases of BSI. Among Gram-negative pathogens, Enterobacteriaceae and Salmonella species were involved in community-acquired infections, whereas more antibiotic-resistant species were also isolated in nosocomial BSI. Among the 13 episodes of Gram-negative hospital-acquired BSI, most were catheter related (10 episodes). Two episodes were classified as primary BSI, and 1 episode was secondary to a digestive tract source.

Treatment and Course

Based on susceptibility results, antibiotic therapy was considered adequate in most cases (92%). After the initiation of antibiotics, the median time until the resolution of fever was 3 days. The median duration of antibiotic therapy was 14 days. Some patients (16%) required surgery to drain an abscess or remove an infected hip prosthesis.

Patients often had a severe infection: 25 (44%) were hospitalized in the intensive care unit. Septic shock occurred in 10 patients (18%). Patients with septic shock were more likely than those without to be further immunosuppressed (40% versus 10.8%; p = 0.05) and to die (50% versus 0%; p = 0.004). The crude hospital mortality rate was 11%. Three of the 5 patients who died were further immunosuppressed. Pathogens responsible for death were Str. pneumoniae (2 patients), and S. aureus, Klebsiella pneumoniae, and Pseudomonas aeruginosa (1 patient each).

Associated Bone-Joint Infection

The mean length of follow-up among survivors was 22.4 ± 20.2 months. Associated bone-joint infection was diagnosed in 18 patients (32% of episodes), including osteomyelitis (9 patients), hip prosthesis infection (5 patients), and arthritis (4 patients). Among patients who had bone-joint infection, the median time to diagnosis was 3 days. Bone-joint infection was diagnosed during the initial BSI episode in 13 patients and 1-6 months after its resolution in 5 patients. The same microorganism previously isolated in blood culture was recovered from a bone or joint specimen in these cases. The pathogens responsible for late bone-joint infection and day of diagnosis after initial admission for BSI were Salmonella species, days 27 and 30; K. pneumoniae, day 60; S. aureus, day 90; and P. aeruginosa, day 180. Duration of antibiotic treatment of BSI did not differ between these 5 patients (mean, 15 ± 7.4 d) and those without bone-joint infection (mean, 14.8 ± 6.9 d).

Comparison of Patients With and Without Bone-Joint Infection

Because of the high incidence of bone-joint infection noted in this series, we compared the baseline characteristics, type of pathogens, biological results, and clinical course between patients with and without bone-joint infection (Table 4). The main factors significantly associated with occurrence of bone-joint infection were a history of osteonecrosis (p = 0.01; relative risk [RR], 2.5; 95% CI, 1.2-5.3) and S. aureus infection (p < 0.001; RR, 3.8; 95% CI, 1.8-8.4). Patients with bone-joint infection also had symptoms (fever or bone pain) longer before admission (p < 0.001). All these factors were independently related to bone-joint infection when tested in a logistic regression analysis (data not shown). A higher CRP concentration was noted in the bone-joint infection group. However, we could not determine a clinically relevant cutoff concentration of CRP that could predict occurrence of bone-joint infection.

TABLE 4: Risk Factors for Bone-Joint Infection (BJI)

DISCUSSION

The first aim of this study was to determine the characteristics and outcome of BSI in adults with SCD. To our knowledge, the current study is the first to focus on BSI in an SCD cohort composed exclusively of adult patients. Most previous studies of BSI in SCD have been conducted in young children, demonstrating an incidence of BSI close to 8 episodes per 100 patient-years²⁸. In children younger than 6 years, Str. pneumoniae accounted for 67% of BSI. The mortality rate was up to 27% in those children who received neither penicillin prophylaxis nor pneumococcal vaccine^9,12,25,28. A study performed in Jamaica found that Str. pneumoniae and H. influenzae were common before 5 years of age, while Salmonella species and Escherichia coli predominated in older children¹⁵.

In contrast, our study revealed several differences. First, the incidence of BSI was much lower in adults (1.2 episodes per 100 patient-years). This is similar to that observed in the adult general hospital population in France (0.98 episodes per 100 patient-years)⁴. An additional cause of immunosuppression, such as HIV infection or corticosteroid use, was noted in 10.7% of our patients and seems to be an important risk factor for BSI and fatal outcome. Second, the fatality rate was low (11%) compared to that observed in a prospective survey of BSI in adults in France (25%) and more recently in the United States (18%)^4,16. The mortality rate was even lower (5%) when we excluded patients with additional immunosuppression. Possible explanations for a better prognosis of BSI in adults with SCD as opposed to the general population include a younger age, which is known to increase the chance of survival after bacteremia in adults⁴; a high proportion of catheter-related infections, which are associated with a better prognosis than other sources of BSI; rapid access to healthcare delivery system; and a high appropriateness of antimicrobial therapy. Third, Str. pneumoniae was rarely responsible for BSI in adults with SCD, accounting for only 10.7% of the episodes. Indeed, we have previously shown that severe pneumococcal sepsis is rare in adults with SCD, except in those with additional immunosuppression, such as HIV infection¹⁰. Although rare, Str. pneumoniae was associated with a high mortality rate in our series since it accounted for 2 of the 5 sepsis-related deaths. Concerning pneumococcal prophylaxis, comparing bacteremic patients with those followed in the cohort showed that the former were less likely to have receive the pneumococcal vaccine during the past 5 years. This emphasizes the need for a greater use of pneumococcal vaccination in this population.

Unlike in children with SCD, the microorganisms most commonly recovered in our series of BSI, whether community- or hospital-acquired, were S. aureus and E. coli. Interestingly, E. coli was almost 3-fold more prevalent than Salmonella species. The only Gram-positive species responsible for nosocomial infection was S. aureus. Whereas meningitis and pneumonia are the main sources of BSI in children, only 3 episodes of pneumonia and 2 episodes of meningitis were recorded in our study (11% of all sources identified)²⁸. Consistent with the predominant role of E. coli in community-acquired BSI, the main sources were the urinary or digestive tracts, representing 64% of the identified sources of community-acquired BSI.

A high proportion (46%) of episodes was hospital-acquired in the current series. This may be due to the management of VOC in these patients, which frequently requires hospitalization and parenteral treatment. Indeed, intravenous catheters were the main source of BSI (41% of episodes). The main pathogen responsible for catheter-related BSI was S. aureus, but Gram-negative bacilli such as Acinetobacter species and P. aeruginosa were also isolated from long-term central venous devices. In contrast, intravascular catheters only accounted for 11% of cases of BSI in the general hospital population⁴. Previous studies have reported that SCD patients have a high risk of central venous device infection, with an incidence rate ranging from 0.46 to 5.5 per 1000 catheter-days^{1,13,17,23,27}. Our experience with 35 patients who had long-term venous catheters demonstrates that device-related BSI affected 77% of the study population. Similar results were reported in a study of 21 patients with 30 devices, which reported a 57% infectious complication rate²⁷. The rate of 1.5 infections per 1000 catheter-days was 4-9 times that of the population of patients with malignancy. Similar to our results, S. aureus was the most common pathogen (59%). Catheter-related BSI was associated with a significant morbidity in this study, because all cases of infection required device removal. Of note, 1 patient died of sepsis resulting from an infected port. In our study, short-term peripheral catheters were the main cause of nosocomial BSI. The fact that the duration of previous intravenous therapy for VOC was longer in patients with catheter-related BSI highlights the need for careful management of intravenous therapy in SCD patients, even for short-term catheters. Appropriate cleansing technique for port access by the caregiver is also mandatory²⁷.

Comparing baseline characteristics of SCD in patients with and without BSI, we found that several variables were associated with an increased risk of sepsis. Hb-S genotype, number of admissions for VOC, and use of implantable venous devices were more frequently found in BSI patients. We also noted that some modalities of treating severe SCD, such as long-term transfusion and intravenous desferrioxamine, were risk factors for the developing sepsis. In fact, these therapeutic modalities could effectively reflect the severity of SCD in these patients. On the other hand, BSI is a rare but well-known complication of red blood cell transfusion^14,21. Similarly, desferrioxamine use is associated with a high frequency of invasive infections with Enterobacteriaceae such as Klebsiella species⁵.

Finally, our most important observation was the particularly high rate of associated bone-joint infection, which was diagnosed in one-third of the episodes. This occurred despite adequate duration of antibiotic therapy of BSI and removal of all infected central venous catheters. Two significant risk factors of associated bone-joint infection were found. First, bacteremic patients with a previous history of osteonecrosis had a 2.5-fold higher risk of bone-joint infection. This could be explained by previous damage to bone tissue during VOC, as osteonecrosis is a typical complication of SCD in adults¹¹. Indeed, x-ray or magnetic resonance imaging revealed a history of symptomatic osteonecrosis in 18 patients (38%). Secondly, the nature of the causative pathogen was an important risk factor of bone-joint infection. In contrast to E. coli sepsis, which was never complicated by bone-joint infection, S. aureus was associated with a 3.8-fold higher risk of bone-joint infection. A similar trend was also noted for Salmonella species but it did not reach statistical significance because of the low number of episodes observed in our study. Our findings are consistent with those observed in a previous study of BSI in children with SCD, which also found a good correlation between the pathogen species (Salmonella and S. aureus isolates) and the occurrence of bone-joint infection¹⁹. Cases of osteomyelitis that occurred late after a BSI episode have been previously reported in patients with SCD³. Although bone-joint infection was mostly diagnosed during the initial course of BSI, 27% of cases of bone-joint infection were diagnosed between 1 and 6 months after resolution of BSI. A prolonged interval between BSI and diagnosis of osteomyelitis has been previously described in a series of 21 patients⁷. Therefore, we recommend being particularly aware of the risk of bone-joint infection. Bacteremic adults with SCD must receive antibiotic therapy for an appropriate duration and must be closely followed-up during the next 6 months, especially when the causative agent is S. aureus or Salmonella species.

In summary, BSI in adults with SCD is a different entity from that observed in children. It is mainly observed in patients with severe underlying disease. Among causative micro organisms, S. aureus predominate over pneumococci. A high proportion of BSI episodes is hospital-acquired, and intravenous catheters are a major source. Careful management of intravenous therapy of VOC is required for the prevention of theses episodes. Clinicians should be aware that a close follow-up is required after resolution of BSI, because of the high-risk for bone-joint infection.