Epidemiology, aetiology and clinical characteristics of clostridial bacteraemia: a 6-year population-based observational study of 386 patients

This is a retrospective population-based observational study on clostridial bacteraemia in the Skåne region, Sweden, between January 2014 and December 2019.

Skåne is a region in the southernmost part of Sweden, with 1.4 million inhabitants (2019). There are ten hospitals dispersed throughout the Skåne region, nine public and one privately run, all using the same electronical medical system. Almost all aspects of modern in-patient healthcare are provided by these hospitals. Thus, the absolute majority of the population of Skåne would seek healthcare in the region.

Case-finding was performed retrospectively through the laboratory databases of Department of Clinical Microbiology, Region Skåne, Lund, Sweden, the sole service-provider of microbiological diagnostics for the region. All blood cultures with a positive finding within the Clostridium genus were extracted from the electronical laboratory database in January 2020. Individuals were identified by their unique personal identity number and in case of multiple bacteraemic episodes with Clostridium, a 3-month deduplication period was employed.

Regional policy during this time period stated that two complete sets (aerobic/anaerobic bottles) should be drawn from separate venepunctures when a bloodstream infection is suspected. From January 1st 2014 to December 31st 2014, the BacT/ALERT® blood culture system was used with the bottles BacT/ALERT FA Plus and BacT/ALERT FN Plus (bioMérieux, Inc.). From January 1st 2015, the blood culture system was changed to BACTEC™ FX with the bottles BACTEC Plus Aerobic, and BACTEC Lytic Anaerobic or BACTEC Plus Anaerobic (BectonDickinson) depending on what bottles were available. Cultures were incubated for 120 h before regarded as negative. The five largest hospitals in the region have on-site blood culturing systems for local incubation, and upon positive signalling, the bottle was transferred to the Department of Clinical Microbiology where species determination was performed with Microflex matrix–assisted laser desorption ionisation-time of flight mass spectrometry (MALDI-TOF MS) (Bruker Daltronics), using the software MALDI Biotyper (MBT) Compass 4.1 with reference database MBT Compass Library DB-7854, according to the manufacturer’s instructions. Several species previously belonging to the Clostridium genus have undergone re-classification since January 2020 as defined in the database “List of Prokaryotic names with Standing in Nomenclature” (LPSN) as of 1 March 2022 [21]. These re-classified species were also included in this study since more changes of taxonomy are to be expected based on phylogenetic studies [22]. However, C. difficile was not included since the new name was already established in 2020. A MALDI Biotyper score of at least 2.0 was required for species determination, and if lower, the 16S rRNA gene was sequenced to assign the species [23]. Isolates only reported at the genus level after neither MALDI-TOF MS nor 16S rRNA methods successfully determined the species were named Clostridium spp. undefined in this study. The minimum inhibitory concentration (MIC) was determined using gradient strips (E-test, BioMérieux) on fastidious anaerobe agar (FAA, LabM) with the European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints for C. perfringens as of 2022 [24]. For the other clostridial species, no breakpoints have been established by EUCAST but guidance is provided for MICs, above which a suggestion is made to advice against the use of that agent [25]. Antibiotic susceptibility testing was not routinely performed for vancomycin or cephalosporins.

Polymicrobial bacteraemia was defined as growth of more than one bacterial species in the same blood culture (2 pairs of anaerobic + aerobic bottles). In cases where multiple clostridial species were identified in the same sample, the most frequently found species was defined as the main species and the least common one as co-pathogen.

Co-morbidities were evaluated using the modified Charlson Comorbidity Index (CCI) [26]. Immunosuppression was defined as an underlying diagnosis of immunodeficiency, any haematological malignancy, corticosteroid treatment (corresponding to ≥ 10 mg prednisolone for > 7 days), other immunosuppressive therapy or ongoing/recent (6 months) chemotherapy or stem cell transplantation within 12 months.

The source of infection was either based on the ICD-10 diagnose code or an obvious diagnosis based on review of the medical records, or otherwise recorded as “unknown”. An abdominal tumour was considered a probable source of infection if no other obvious source was present. A nosocomial infection was defined as the first positive blood culture drawn ≥ 48 h of hospital admission.

All clinical characteristics were registered within 24 h of blood culture. Fever was defined as a temperature > 38.0 °C and hypotension as a mean arterial pressure ≤ 70 mmHg. Thrombocytopenia was defined below 150 × 10⁹/L, without previous known thrombocytopenia. Acute kidney failure was defined as a creatinine above reference, without previous known kidney failure. Sepsis and septic shock were defined according to the Sepsis-3 criteria [27]. Massive haemolysis was considered present when the laboratory reported that the sample could not be analysed due to haemolysis. Admission to the intensive care unit (ICU) was recorded within 10 days from the first positive blood culture. Treatment with a surgical procedure included all invasive therapeutic procedures performed as infection source control during the hospital stay after obtaining the first positive blood culture. All-cause mortality was recorded within 28 days from the first positive blood culture.

The enumerator for incidence calculations was the number of unique infectious episodes, as defined above. The population denominator was defined as the year-end population with residency within the Skåne region, stratified by age and sex, retrieved from the database of Statistics Sweden, a governmental agency providing population statistics [28]. The age-standardised incidence rate was estimated by applying the direct standardisation method to the 2013 European standard population (in 5-year age strata, truncated at 90 +). Incidence rate ratios (IRR), including 95% confidence intervals, were used for comparisons across age and sex strata. Associations with 28-day mortality were evaluated using bivariate and multivariate logistic regressions, presenting odds ratios (OR) with 95% confidence intervals. In multivariate analysis, variables with a high degree of missingness (> 5%) were excluded, as were variables with suspected collinearity. To provide scale-independent comparability, ORs were presented for the 3rd quartile vs 1st quartile for continuous variables.

During the study period, 386 unique episodes with clostridial blood-stream infections were identified. The crude (95% CI) incidence rate was 4.8 (4.4–5.3) per 100.000 person-years and the age-standardised incidence rate was 4.9 (4.5–5.5). The median age was 76 years (IQR 66–83) and 217/386 (56%) of patients were males. There was a substantial age effect, with an IRR (95% CI) of 34.3 (25.5–46.6) for those aged 80 + vs 0–59, see Fig. 1. There was also a predominance for males in all age strata, with an overall crude incidence rate for males of 5.4 (4.8–6.2) per 100.000 person-years vs 4.2 (3.6–4.9) for females, resulting in an IRR (95% CI) of 1.29 (1.06–1.58). Incidence rates were stable over time during the 6-year period. When the 2022 taxonomy was used, the number of unique episodes was 322, resulting in an age-standardised incidence rate of 4.1 (3.7–4.6).

The most common species was C. perfringens accounting for 156/386 (40%) isolates, followed by C. septicum 52 (13%) and C. ramosum 48 (12%). The species distribution is presented in Table 1. Polymicrobial bacteraemia was present in 194/386 (50%) of cases, most commonly Enterobacterales (39%), followed by Bacteroides (14%) and Enterococcus (13%). In a majority, only one co-pathogen was present (124/194, 64%). Differentiating from the other species, bacteraemia with C. septicum was monomicrobial in 96% (50/52) and with C. symbiosum in 89% (8/9) of the cases.

Using the recently published EUCAST MIC breakpoints for C. perfringens, all but 1, 1 and 3/156 C. perfringens isolates were susceptible to penicillin (MIC ≤ 0.5), metronidazole (MIC ≤ 4) and piperacillin/tazobactam (MIC ≤ 0.5), respectively, but 92/156 (59%) were resistant to clindamycin (MIC > 0.5). Due to the lack of clinical breakpoints for non-perfringens species, the frequency of susceptible isolates cannot be determined. In general, the MIC distributions varied considerably between species, but were almost universally low (MIC ≤ 4) for metronidazole. The MICs for all four antibiotics were also generally low (corresponding to S for C. perfringens) for C. septicum (Fig. 2) and MICs for carbapenems were low (≤ 0.125) for all species (not shown).

The median CCI was 2 (IQR, 1–5). Of all patients, 84 (22%) had no underlying co-morbidities according to the CCI scoring system of which 34 (40%) had a diagnosis of either appendicitis or cholecystitis/cholangitis. C. septicum was unusual in the group without co-morbidities, accounting for only 4/84 (5%). Malignancy was the most common co-morbidity, in 182 (47%) patients, most pronounced for C. septicum in 40/52 cases (77%). Of the malignancies, haematological malignancies accounted for 26 (14%) of cases and 81 (45%) had metastatic disease.

Clinical information was available for 384 patients. Intraabdominal infection sites predominated where cholecystitis and cholangitis were the most common diagnoses in 49/384 cases (13%), caused by C. perfringens in 40/49 (82%). Soft tissue infections accounted for 25 (7%) of which two cases of necrotizing fasciitis and one case of Fournier’s gangrene were identified. No case of gas gangrene was identified (Table 2).

In most cases, symptoms appeared within 24 h before the first blood cultures were obtained (median 0.5 days, IQR 0–2). In total, 54% of patients had fever, 37% had hypotension and 69% and 17% met the criteria of sepsis and septic shock, respectively (Table 3).

Information on empiric antibiotic treatment was available for 385 patients of whom 158 (41%) received piperacillin/tazobactam, 114 (30%) received cefotaxime, in combination with metronidazole in 15 cases, and 51 (13%) received a carbapenem. Forty-two patients (11%) did not receive any empiric antibiotic treatment. Surgery was considered necessary as source control in 155 patients (40%), of whom 133 (35%) underwent surgical intervention. ICU treatment was given to 72 (19%) patients, see Table 3.

The median length of stay at the hospital was 10 days and the 28-day mortality rate was 26% (100/384). Sixty-seven of the 100 patients who died did not receive intensive care. Bivariate analysis revealed that age, absence of fever, high C-reactive protein (CRP) level on arrival, high SOFA-score, elevated lactate and the presence of sepsis or septic shock within 24 h from obtaining the first blood culture were all significantly associated with mortality. A similar pattern was seen in multivariate analysis (Table 4). Polymicrobial and monomicrobial infections differed slightly in clinical presentation and severity, but the difference in mortality was not statistically significant (Table 4 and Supplementary Table). The mortality in the group treated with cefotaxime only did not differ from overall mortality (24% vs 26%) (data not shown).