Background

Invasive infection with Listeria monocytogenes is a serious illness characterized by up to 46% mortality [1,2,3,4,5,6,7]. Currently, the recommended therapy is an aminopenicillin with or without the addition of gentamicin (“combination therapy”) [8]. While the choice of an aminopenicillin is supported by clinical experience of treatment success [9,10,11], the addition of gentamicin is based on in vitro studies which showed synergism between the two antibiotics [12,13,14,15,16,17]. Results of in vivo studies have been contradictory regarding the superiority of combination to therapy with ampicillin alone (“monotherapy”) [18,19,20,21].

Due to the rarity of human listeriosis, no randomized controlled trials have been carried out evaluating efficacy of different antibiotic regimens and available data come from observational studies. Among studies addressing combination therapy and mortality [1, 22,23,24,25,26,27,28,29], many reported outcomes exclusively in patients with central nervous system (CNS) infection [22, 24, 27, 28]. The results were not consistent, with some studies indicating a lack of benefit or even harm resulting from combination therapy [22, 23, 25,26,27,28,29] and others finding an association with decreased mortality [1, 24]. One factor which may possibly explain some of this variation is the timing of therapy. As listeriosis is a rare entity, patients seldom receive empiric combination therapy, and gentamicin is typically added only after initial results of microbiologic testing. Delayed initiation of an aminoglycoside could decrease the utility of combination therapy and bias results towards lack of efficacy.

The incidence of listeriosis in Israel is higher than that reported by other developed countries [3, 30, 31], and the need to address the issue of optimal therapy is pressing. The objective of this study is to evaluate the efficacy of antibiotic therapy on invasive L. monocytogenes infections while adjusting for the timing of treatment onset.

Methods

Study design

This was a retrospective cohort study of patients who were hospitalized in eleven acute care medical centers in Israel between 1 January 2008 and 31 December 2014. The study included 10,128 (60.9%) out of the total of 16,644 admission beds in Israel [32] and included six tertiary centers and five secondary medical centers. The study was approved by the ethics committees of all participating institutions.

Participants

We included patients aged 18 years and older in whom L. monocytogenes was cultured from sterile sites with clinical evidence of systemic infection or CNS involvement, including meningitis or brain abscess. Pregnant women were excluded. Patients were identified by searching the records of clinical microbiology labs in the participating hospitals. Each patient was included once only; for patients with multiple episodes of infection, data were gathered for the first episode only. Patients who died within 48 h of culture-taken date (CTD) were included in the description of demographic and clinical features but excluded from the outcome analyses on the assumption that early mortality was due to factors unassociated with therapy.

Variables

The primary outcome was all-cause 30-day mortality. Secondary analyses were planned for the primary outcome in patients with primary bacteremia and CNS involvement, respectively. The exposure was antibiotic therapy with monotherapy, early combination therapy, or others. Monotherapy was defined as treatment with an antilisterial penicillin (including ampicillin, penicillin, piperacillin, amoxicillin-clavulanate, ampicillin-sulbactam, or piperacillin-tazobactam) for at least 7 consecutive days subsequent to the culture-taken date, without concomitant effective antibiotics (including aminoglycosides) for more than 3 consecutive days. Early combination therapy was defined as treatment for at least 7 consecutive days subsequent to CTD with an antilisterial beta lactam and gentamicin; the latter initiated within 48 h of culture results. Information on potential confounders was gathered including data on patient demographics, background medical conditions, sepsis severity on CTD, and antibiotic therapy prior to admission. The information was obtained retrospectively from patients’ written and electronic records.

Statistical analysis

We compared patients with L. monocytogenes infection who received monotherapy, early combination therapy or other therapy, and a 30-day mortality. Categorical variables were compared using the χ2 test and continuous variables using the t test or Mann-Whitney U test according to their distribution. Confounders and other risk factors for mortality found significant on univariate analysis and noncorrelated were entered into a binomial logistic regression model. To further control for the effects of confounding, a propensity score method with stratification was used. A propensity score for receiving early combination therapy was calculated by running a logistic regression using variables which were found to be associated with combination therapy on univariate analysis. The propensity score was divided into four equally sized groups using quartiles, and a conditional logistic regression for a 30-day mortality was then performed stratifying on these quartiles. Odds ratios (ORs) with 95% confidence intervals (CIs) are reported. Analyses were conducted using IBM SPSS version 25 software.

Results

One hundred ninety-three nonpregnant adults with invasive listeriosis were identified; most patients were women (108, 56.0%) and fully functional (143, 74.1%). The median age was 69 (60–81) years. The median duration of illness prior to CTD was 2 (1–3) days, and most (162, 83.7%) patients did not receive antibiotics prior to CTD. Sixty-three (32.6%) patients had CNS involvement, 124 (64.2%) had primary bacteremia, and 6 (3.1%) had other diagnoses. Fifty-nine (30.6%) patients were treated with early combination therapy, 90 (46.6%) received monotherapy, and 44 (22.8%) received other treatments (Fig. 1). Factors associated with early combination therapy included fully functional status and CNS involvement, while factors associated with monotherapy included Arab ethnicity and primary bacteremia. Further information is available in Table 1.

Fig. 1
figure 1

Patient flow diagram. AAT—appropriate antibiotic therapy; CRD—culture result date; TMP/SFX—trimethoprim/sulfamethoxazole

Full size image
Table 1 Factors associated with treatment
Full size table

A total of 190 patients were available for the univariate and multiple-variable analyses after excluding early mortality. Overall 30-day mortality was 20.5% (39/190); mortality among patients with CNS involvement was 21.0% (13/62) and 24.4% (30/123) in patients older than 65 years. Factors associated with survival included higher baseline functional capacity. Factors associated with a 30-day mortality included congestive heart failure (CHF), higher sequential organ failure assessment (SOFA) score on CTD, and elevation of liver function tests (LFTs) greater than two times the upper limit of normal on CTD (Table 2). Source of infection, treatment type, and time from CTD to initiation of appropriate therapy were not associated with mortality.

Table 2 Univariate analysis, 30-day mortality
Full size table

The multiple-variable analysis included all of the factors which were significantly associated with mortality except elevated LFTs which correlated with SOFA score. Additionally, the exposure variable was forced into the model. Factors associated with greater 30-day mortality in multiple-variable analysis included increasing SOFA score (OR 1.186, 95% CI 1.085–1.296) and treatment other than early combination therapy or monotherapy (OR 3.899, 95% CI 1.228–12.382) (Table 3). Fully functional status at baseline was associated with lower mortality (OR 0.343, 95% CI 0.153–0.768). Monotherapy had a nonsignificant trend to higher 30-day mortality compared with early combination therapy (OR 1.947, 95%CI 0.691–5.487). After stratifying by the propensity score, increasing SOFA score (OR 1.136, 95% CI 1.075–1.201) and treatment other than early combination therapy or monotherapy (OR 3.930, 95% CI 1.407–10.977) remained associated with mortality (Table 4). The trend toward higher mortality with monotherapy compared with early combination therapy remained nonsignificant (OR 2.312, 95% CI 0.915–5.843). Results were similar in separate analyses of patients with primary bacteremia and CNS involvement (Supplementary materials). Length of beta lactam-aminoglycoside treatment was not independently associated with mortality when included in the regression (OR 0.666 for > 3 days vs. 0, 95% CI 0.140–3.161).

Table 3 Binomial regression
Full size table
Table 4 Conditional regression stratified by propensity score
Full size table

Discussion

In a retrospective cohort of patients with invasive listeriosis, overall 30-day mortality was 20.5%. No difference was observed in mortality between patients who received a penicillin-based monotherapy or combination therapy with gentamicin initiated within 48 h of culture result after adjusting for other mortality risk factors. Treatment other than penicillin-based monotherapy or early penicillin-aminoglycoside combination therapy was associated with higher mortality.

The use of beta lactam-aminoglycoside combination therapy for infections based on in vitro evidence of synergy has come into question in recent years [33,34,35,36,37]. In light of these data, there has been a reevaluation of the efficacy of combination therapy in patients with listeriosis. More recent in vitro data has shown that the addition of gentamicin can lead to persistence of intracellular infection and promote bacterial survival [38]. Research on pregnancy-associated disease indicates that gentamicin may increase dissemination of listeria from the placenta to other organs [39]. Clinical data from several retrospective cohort studies [22, 23, 25,26,27,28] also failed to support the addition of an aminoglycoside and in some cases indicated increased mortality with combination therapy. However, small sample size and a lack of adjustment to confounders limit the generalizability of their results.

The largest study to date on invasive listeriosis (MONALISA) evaluated a prospective cohort of 818 patients for a 3-month mortality [1]. They found that treatment with gentamicin was associated with lower mortality among patients with CNS involvement or non-CNS, nonmaternal-fetal bacteremia (OR 0.60, 95% CI 0.38–0.94), and that amoxicillin-gentamicin combination therapy given for greater than three days was independently associated with lower mortality compared with monotherapy (OR 0.35, 95% CI 0.22–0.56). However, the adjusted analysis did not include bacteremia/CNS infection as a variable although this was an important confounder in the cohort (as in ours) given that significantly more patients with CNS infection received combination therapy (192/252 vs. 170/427, p < 0.001) while experiencing a lower rate of a 3-month mortality (72/252 vs. 194/427, p < 0.001). This raises the possibility that part of the advantage observed with combination therapy was due to confounding by clinical presentation. In addition, no information was presented on timing of appropriate therapy vis-à-vis symptom onset or diagnosis. In our cohort, no association was seen between time from symptom onset to presentation or time from CTD to onset of appropriate therapy and mortality; of the other studies published to date, two found an association between mortality and appropriate empiric therapy/time to appropriate therapy and two did not [23, 26,27,28]. Thus, questions remain about the role of aminoglycosides in the treatment of listeriosis, especially when there is a delay in initiation.

The 30-day mortality (20.5%) observed in our study was lower than that seen in MONALISA for bacteremic patients but similar to rates from previous retrospective studies [22,23,24,25,26,27,28], including a meta-analysis [2]. One reason for the differences might be variability in patient populations. The strongest predictors of mortality in our analysis were SOFA score at onset and baseline functional status; no sepsis severity scores were available for patients in the MONALISA trial, and functional status was not reported, making a comparison difficult. Additionally, differences in healthcare infrastructure may have played a role. Six of the hospitals included in our study were tertiary-level and the remaining five were secondary, with no community-level centers; all hospitals involved were university-affiliated. Notably, there was no significant difference in length of symptoms prior to diagnosis between our study and the MONALISA trial, implying that the lower mortality which we observed was not due to survival bias.

Our analysis had a number of limitations. First, although multicenter and inclusive of most hospital beds nationwide, the cohort size was relatively small and did not cover the entire country (11/28 hospitals nationwide). The MONALISA study that found an advantage to combination therapy included more than 800 patients with a mortality difference of 17% at 3 months; using the mortality rate we observed and assuming the odds ratio for the 30-day mortality we calculated is true, a sample size of 408 would be necessary to demonstrate significance of α = 0.05. Second, this was a retrospective observational study and it is likely that potential confounders for choice of antibiotic treatment were not accounted for in our analysis. Third, in our definitions of antilisterial beta lactams, we included penicillin, ampicillin/amoxicillin, and piperacillin, assuming equal effectiveness. While evidence exists for the equivalence of penicillin and ampicillin [40], data comparing ampicillin with piperacillin are scarce. The mean inhibitory concentration (MIC) against L. monocytogenes is higher for piperacillin than ampicillin [41, 42]; however, the clinical significance of this fact is uncertain; in the only study to compare piperacillin with other antilisterial penicillins, only five patients received piperacillin as definitive therapy of whom three survived [26]. Given the widespread use of piperacillin-tazobactam as empiric therapy for sepsis in immunosuppressed patients, further investigation of its clinical efficacy in listeriosis is advisable.

In conclusion, we did not find an association between combination therapy and the 30-day mortality, although associations tended in favor of combination therapy. The optimal regimen for treatment of listeriosis remains unclear. Current global interconnectivity might allow for multicenter investigator-initiated studies of rare infections; the role of combination therapy will be best assessed in randomized controlled trials.