Confirmed Bacillus anthracis Infection among Persons Who Inject Drugs, Scotland, 2009–2010

Patients who died had an increased sequential organ failure assessment score and need for vasopressors.

In Scotland, the 2009 outbreak of Bacillus anthracis infection among persons who inject drugs resulted in a 28% death rate. To compare nonsurvivors and survivors, we obtained data on 11 nonsurvivors and 16 survivors. Time from B. anthracis exposure to symptoms or hospitalization and skin and limb findings at presentation did not differ between nonsurvivors and survivors. Proportionately more nonsurvivors had histories of excessive alcohol use (p = 0.05) and required vasopressors and/or mechanical ventilation (p<0.01 for each individually). Nonsurvivors also had higher sequential organ failure assessment scores (mean + SEM) (7.3 + 0.9 vs. 1.2 + 0.4, p<0.0001). Antibacterial drug administration, surgery, and anthrax polyclonal immune globulin treatments did not differ between nonsurvivors and survivors. Of the 14 patients who required vasopressors during hospitalization, 11 died. Sequential organ failure assessment score or vasopressor requirement during hospitalization might identify patients with injectional anthrax for whom limited adjunctive therapies might be beneficial.

B acillus anthracis infection in humans has typically been
classified as cutaneous, gastrointestinal, or inhalational on the basis of the bacterium's route of entry (1). However, in Scotland, United Kingdom, during 2009-2010, a total of 47 patients had confirmed B. anthracis soft tissue infection related to injection of contaminated heroin (2,3). This form of B. anthracis infection appears to be distinct from cutaneous disease and has been termed "injectional" anthrax (2)(3)(4)(5). In addition to confirmed cases, 35 probable and 37 possible cases in Scotland, 5 confirmed cases in England, and 2 confirmed cases in Germany also were identified. This initial outbreak ended in late 2010, but since the summer of 2012, new cases have been reported in the United Kingdom and Europe (3,4). Although 1 case of injectional anthrax was recognized in Norway in 2001, the patients in 2009-2010 constitute the first large outbreak of this newly recognized and poorly characterized form of anthrax (5).
Health Protection Scotland (HPS) has published epidemiologic analyses of the 2009-2010 outbreak (3,6,7). Among other findings, analysis suggested associations between longer injecting histories, opioid substitution therapy, and alcohol use and risk for B. anthracis infection in persons who inject drugs (PWID) (6). Several case reports from the outbreak also have been published (8)(9)(10)(11)(12), but they did not include systematic examinations of the physical, laboratory, and surgical findings or of therapies administered. Notably, although 13 of the 47 persons from Scotland who had confirmed cases died, no published report has compared findings in survivors and nonsurvivors. Such a comparison is needed for the prognosis and management of future cases. We therefore sent a questionnaire regarding these issues to clinicians who had treated PWIDs in whom B. anthracis infection was confirmed in Scotland during the outbreak.

Approval
We used data collected during routine hospital care of patients. Patient identifiers were removed from data before analysis. Because of the retrospective nature of the study and the anonymity of data, the West of Scotland Research Ethics Service (Glasgow, Scotland, UK) and the Office of Human Subjects Research from the Clinical Center at the National Institutes of Health (Bethesda, MD, USA) exempted the study from formal review.

Data Collection
We developed an electronic questionnaire that requested information in several areas. These were: general information (i.e., age, sex, and medical and drug histories); current illness; data at the time the patient sought care, including skin and limb findings, vital signs, laboratory findings, and diagnosis (i.e., was anthrax infection or sepsis initially suspected?); medical and surgical treatments at the time the patient sought care or later (including antibacterial drugs; need for hemodynamic, respiratory, renal replacement, or blood product support; and use of anthrax immune globulin, a polyclonal antibody produced by Cangene [Winnipeg, MB, Canada] and made available by the Centers for Diseases Control and Prevention [Atlanta, GA, USA]; microbiological data supporting the diagnosis of anthrax infection and the time at which the diagnosis was confirmed; surgical findings; other procedures performed during hospitalization; levels of organ injury based on the sequential organ failure assessment (SOFA) score; and outcomes, including survival, time in the intensive care unit (ICU), and total time in hospital.
During this anthrax outbreak, HPS identified 13 hospitals to which the 47 patients with confirmed anthrax were admitted (3). In March 2012, two of the authors (L.D. and M.B.), who were members of the HPS anthrax outbreak control team, asked physicians known to have treated persons with confirmed anthrax to complete the questionnaire. The questionnaire was sent to these physicians in early April 2012. In June 2012, physicians who had not yet returned it were asked to do so. Data from all questionnaires completed by the end of August 2012 were analyzed. We received no additional questionnaires after August 2012. Contributors were subsequently contacted to clarify missing or unclear responses.

Data Analysis
A variable was reported on only if >50% (i.e., >14) of questionnaires provided definitive data. SOFA score was calculated as previously described for each patient for whom data were obtained within the initial 24 hours after they sought care (13). Measures of time were expressed as medians (interquartile range [IQR]). We used Wilcoxon rank sum test to compare these times between survivors and nonsurvivors. Categorical data (i.e., chief complaints, types of initial surgery and gross tissue findings, skin and limb findings, and treatments) were analyzed with Fisher exact or χ 2 test, where applicable, and continuous data (i.e., laboratory data, age, vital signs, and SOFA score) were examined with 1-way ANOVA (analysis of variance).
(intravenous or intramuscular) of contaminated drug injection did not differ significantly (p>0.19 for all, data not shown). HIV infection status was available for only 10 patients (4 nonsurvivors), and hepatitis C infection status was available for only 12 patients (5 nonsurvivors). One survivor was HIV positive, and 4 survivors and 5 nonsurvivors had histories of hepatitis C virus infection. Although the questionnaire requested other medical history (e.g., hepatitis B virus infection status), these data were not provided in sufficient numbers for analysis.
At the time they sought care, nonsurvivors had lower temperatures and systolic blood pressures and higher respiratory rates than survivors (p<0.01) ( Figure 2), but other vital signs (mean + SEM), including diastolic blood pressure (68 + 4 vs. 61 + 5 mm Hg), heart rate (111 + 5 vs. 118 + 8 beats per minute), and capillary refill (2.6 + 0.4 vs. 3.6 + 0.5 seconds) did not differ significantly. Four nonsurvivors, but only 1 survivor, had a temperature <36°C when they sought care, whereas 1 nonsurvivor and 7 survivors had temperatures >38°C. Also, whereas 4 nonsurvivors had systolic blood pressures <90 mm Hg, no survivor did. The Glasgow coma score recorded during the initial 24 hours was lower for nonsurvivors than for survivors (p = 0.008) ( Figure 2). The proportion of survivors and nonsurvivors for whom specific skin and limb findings were available did not differ significantly (p>0.19 for all) ( Table 2).
Nonsurvivors had lower serum sodium, corrected calcium, albumin, and platelet levels and higher bilirubin, percentage circulating neutrophils, hemoglobin concentration, international normalized ratio, prothrombin time, partial thromboplastin time, and C-reactive protein levels (p<0.05) (all shown in Figure 3 except prothrombin time, which was median [IQR] 12 [11][12][13] vs.15 [13.5-16.1]). For a greater proportion of nonsurvivors than survivors, arterial blood gases were measured when they sought care (10 [91%] vs. 6 [38%], p = 0.008). In patients with for whom the following values were measured, nonsurvivors had lower bicarbonates and greater base deficits (p = 0.02). Other laboratory data did not differ significantly between survivors and nonsurvivors (Table 2). Headache followed by delirium Generalized 3 Pain of right thigh Local 4 Pain of right buttock, chills, malaise Local, generalized 5 Swelling of right arm, seizure Local, generalized 6 Pain, erythema of right groin increasing in area Local 7 Confusion, lethargy, malaise Generalized 8 Pain of right hip, abdominal pain, seizures Local, generalized 9 Abdominal pain, fever Generalized 10 Fainting Fatigue, malaise Generalized *Patients are numbered in the order in which they sought care during the outbreak.

Confirmation of B. anthracis Infection
The basis for the microbiological diagnosis of B. anthracis for the 47 confirmed anthrax cases has been pub-  Table 2). We had insufficient data to determine how often bacteria other than B. anthracis were found in blood or tissue samples.

Hospital Course
Proportionately more nonsurvivors than survivors received ICU care ( Table 4). One survivor had an arm amputated above the elbow, and 4 received skin grafts.
Receipt of packed erythrocytes, fresh frozen plasma, cryoglobulin, and platelets; renal replacement therapy; and pleural or peritoneal drainage did not differ significantly between survivors and nonsurvivors ( Table 2). For 9 (36%) of 25 patients for whom information was available, cardiac function was assessed: echocardiography for 4 patients, troponin measures for 3, and lithium dilution cardiac output and pulse contour cardiac output for 1 each. Of these, 3 nonsurvivors were noted to have evidence of myocardial dysfunction on the basis of echocardiography, lithium dilution cardiac output, or pulse contour cardiac output; 1 survivor had an elevated troponin level. Causes of death for the 7 patients for whom autopsies were reported were as follows; multiple system organ failure caused by B. anthracis sepsis, 2 patients; necrotizing fasciitis related to B. anthracis, 1; sepsis and hemorrhagic meningitis with B. anthracis infection, 2; subarachnoid hemorrhage, 1; and myocardial infarction, 1.

Discussion
Our review of 27 confirmed cases of B. anthracis infection in PWID compares clinical findings in survivors  and nonsurvivors of this newly described form of infection. Although duration of symptoms and time to seeking hospital care did not differ between survivors and nonsurvivors, the severity of illness did. Most survivors reported localized symptoms related to the injection site, and none required vasopressor therapy or mechanical ventilation. In contrast, most nonsurvivors had generalized symptoms and evidence of sepsis, which required both vasopressor support and mechanical ventilation. Nonsurvivors also had lower systolic blood pressures and Glasgow coma scores; higher respiratory rates; worsened base deficits; higher levels of hemoglobin (consistent with hemoconcentration) and C-reactive protein; higher international normalized ratio; and lower sodium and albumin levels and platelet counts. During hospitalization, all nonsurvivors required vasopressor and ICU support, whereas only 3 and 7 survivors, respectively, required these. SOFA scores were substantially higher in nonsurvivors than survivors. Thus, assessing the need for aggressive cardiopulmonary support or determining a score like SOFA for patients with injectional anthrax can help identify those for whom prognosis is particularly poor and more aggressive therapy is needed.
Possibly consistent with prior analysis showing an association between excessive alcohol use and risk for B. anthracis infection in PWID, we found a higher incidence of excessive alcohol use among nonsurvivors than survivors (6). Increased bilirubin and decreased albumin levels in nonsurvivors might in part have reflected preexisting alcoholic liver disease. Although age did not differ significantly between survivors and nonsurvivors, the latter tended to be older, a finding consistent with analysis of inhalational B. anthracis infection (14).
Differences in outcome between survivors and nonsurvivors did not appear related to variation in treatment. All patients received antibacterial drugs from the time they sought care, and the types and numbers of antibacterial drugs administered did not differ. Also, the proportion of patients who had >1 surgeries and the time from admission to initial surgery did not differ. Finally, similar proportions of survivors and nonsurvivors received anthrax immune globulin, and the median time to treatment for these groups did not differ.
The most common skin and limb findings were localized edema, pain, and erythema. Although these findings are consistent with soft tissue infection, their presence did  - not differ between survivors and nonsurvivors (15). Thus, skin and limb findings and even the need for surgery did not appear to predict worsened prognoses. Notably absent in most patients was the eschar formation classically associated with cutaneous B. anthracis infection (16,17). This absence combined, with the frequent need for surgery and the overall high death rate despite receipt of antibacterial drugs, supports the observation that the pathogeneses of injectional and cutaneous anthrax differ (3,(4)(5)(6).
In general, the severity of soft tissue infection and its requirement for surgery varies from mild to severe on the basis of the depth of tissue involvement and tissue necrosis (15). The B. anthracis soft tissue infections reviewed here reflect this range of disease. Two survivors had symptoms consistent with cellulitis, were treated only with antibacterial drugs, and were discharged within 1-2 days. However, 10 survivors and 5 nonsurvivors required debridement, fasciotomy, or laparotomy on >1 occasion; 10 had evidence of tissue necrosis. Even among the 5 survivors who did not have surgery, 3 required hospitalization for >7 days, which suggests severe infection. Without additional data about co-existing conditions, we cannot determine whether outcomes from soft tissue infection in PWID differ between B. anthracis and other bacteria.
B. anthracis has a cell wall that elicits a robust host inflammatory response and the endothelial dysfunction, shock, and organ injury with which this response is associated (1,(18)(19)(20)(21). However, B. anthracis also releases lethal and edema toxins that can produce this same dysfunction but through very different mechanisms than the cell wall  For panels E and F, n = the number of survivors or nonsurvivors for whom SOFA scores were calculated. SOFA, sequential organ failure assessment; ICU, intensive care unit. (22,23). Whether, as a result of these diverse nontoxin and toxin components, the manifestations of soft tissue infection with B. anthracis differ from those of other bacteria is unclear. Several differences between nonsurvivors and survivors, such as reduced systolic blood pressure and sodium and worsened acidosis, are associated with worsened outcome with other types of soft tissue infection (24)(25)(26). However, of patients who required vasopressor treatment, almost 80% died. This death rate is high, even for patients identified with septic shock on the basis of need for vasopressors. However, this finding is consistent with the 2001 US outbreak of inhalational B. anthracis in which all patients in whom shock developed died (27). Also in the current review, nonsurvivors bled more during surgery, possibly because of an increase in international normalized ratio and a decrease in platelets. Although excessive bleeding is not typically associated with soft tissue infection, it is associated with inhalational and gastrointestinal B. anthracis infection (28,29). B. anthracis lethal and edema toxins inhibit components in the innate and adaptive immune responses (22,23). This inhibition might contribute to infection, as well as suppress signs typically associated with an activated host inflammatory response (3). However, although temperature was lower in nonsurvivors than in survivors, circulating leukocyte counts, percentage of neutrophils, and C-reactive protein were higher and in ranges approaching or consistent with invasive soft tissue infection caused by other bacteria (30,31). Whether toxin production interferes with host defense and influences the features and course of injectional B. anthracis infection requires further study.
Several lines of evidence suggest that B. anthracis lethal toxin produces direct cardiac dysfunction (22). Whether such dysfunction contributes to clinical B. anthracis infection is unclear because there are few measures of cardiac function in patients. Although 4 of the 9 patients in the current review were described as having evidence of cardiac dysfunction, these data were limited. More comprehensive investigation of cardiac function is necessary during future outbreaks of B. anthracis.
This study has limitations. First, data were collected >2 years after patients sought care, were not obtainable for some questions on patients we included in the analysis, and were unavailable for 20 of the 47 confirmed cases. However, these 27 patients included 11 of the 14 nonsurvivors from the outbreak and probably were fairly representative of nonsurvivors. Second, data were limited regarding comorbidities, particularly HIV infection and viral hepatitis status, which might have influenced outcome. Information about other co-morbidities (e.g., diabetes, heart disease, and chronic lung disease) might have been informative as well. Third, although 1 survivor and 1 nonsurvivor described headache, using the Glasgow coma scale to assess neurologic status might not have captured other patients with this symptom, a possible manifestation of underlying 0/11 (0) 1/11 (9) 5/9 (56) 7/11 (64) 1/11 (9) 1/10 (10) 1/10 (10) 1 --   (14) *Debride, debridement; thigh ex., thigh exploration; +, type of surgery or tissue finding performed or noted; -, type of surgery or tissue finding not performed or noted; NA, observation not available. †Patients are numbered in the order in which they sought care. ‡No. (%) patients with the type of surgery or finding/total no. patients for whom data were reported. §For proportion of positive responses by survivors vs. nonsurvivors for whom data were reported, p<0.05. meningeal infection. Fourth, comparisons of therapies administered to survivors and nonsurvivors later than when they initially sought care might have been confounded by patients' length of hospitalization; however, data were insufficient to analyze the influence of this variable. Finally, autopsy findings were available only for 7 patients.
The 2009-2010 outbreak of B. anthracis infection among PWID in Scotland was considered over at the end of 2010 (3). However, during June 2012-December 2013, thirteen new cases were reported in the United Kingdom and Europe (4,32). The death rate for these patients has been close to 50%. Findings from the patients in the current review, combined with findings from newer cases, emphasize the need to better understand the pathogenesis and management of this recently identified form of B. anthracis infection.