https://orcid.org/0000-0002-0416-3996
Abstract
We analyzed the impact of a diagnostic meningitis/encephalitis polymerase chain reaction panel. Introducing the test resulted in no significantly shortened empiric antibiotic streptococcal/meningococcal treatment (−1 day [95% confidence interval {CI}, 0–3 days]; P = .16), nor hospitalization (−2 days [95% CI, 0–3 days]; P = .86). Empiric treatment duration against Herpesviridae and Listeria monocytogenes remained unchanged, and pathogen-specific diagnoses were not increased.
Central nervous system infections such as meningitis and encephalitis are associated with high mortality rates and increased long-term morbidity. Therefore, early empiric anti-infective treatment is crucial. Current guidelines suggest initiating empiric intravenous broad-spectrum antibiotics in suspected meningitis cases and adding antivirals in case of encephalitis [1]. Culture-based approaches in cerebrospinal fluid (CSF) diagnostics have a turnaround time of several days [2], potentially leading to prolonged empiric and therefore broader nontargeted coverage. This may result in longer hospital stays, increase the risk of drug-induced toxicity [3], and trigger antimicrobial resistance. While antibiotic stewardship efforts play a significant role in reducing antimicrobial resistance, there has been a shift in focus toward diagnostic stewardship [4]. Diagnostic stewardship aims to steer provider behavior by guiding the diagnostic process, in order to avoid overtreatment.
The introduction of the PCR-based BioFire FilmArray Meningitis/Encephalitis Panel (bioMérieux SA, Marcy-l’Etoile, France) has shortened the time to diagnosis in some cases of meningitis and encephalitis [2]. Previous smaller studies have investigated the effect of this test on antibiotic treatment duration [2, 5, 6], on antiviral treatment duration [6], and on length of hospital stay [5, 7].
However, these studies were limited by small sample sizes and primarily reported mean treatment durations without accounting for censoring due to patient transfers or discharges, nor adjustment.
We therefore conducted a single-center retrospective cohort study at a tertiary care center to assess the influence of the BioFire FilmArray Meningitis/Encephalitis Panel on the above-mentioned parameters, and including an adjusted survival analysis, comparing periods before and after the introduction of the polymerase chain reaction (PCR)–based diagnostic pathway.
METHODS
We retrospectively identified in the hospital laboratory system inpatients >18 years of age evaluated for meningitis and/or encephalitis using CSF at our tertiary care university hospital during the 2 study periods. Patients without CSF pleocytosis (CSF count <5 M/L) were excluded. Cohort A comprised respective patients from October 2013 until September 2016 (prior to introducing the BioFire FilmArray Meningitis/Encephalitis Panel), who thus underwent the traditional evaluation using bacterial CSF culture combined with herpes simplex virus (HSV) 1/2– and varicella zoster virus (VZV)–specific PCR assays. Cohort B included patients from October 2016 to September 2019, after the introduction of the PCR-panel based testing strategy in addition to the culture-based CSF workup. The strategy for further serological tests remained unchanged between the 2 periods: tick-borne encephalitis, human immunodeficiency virus, and syphilis serology; Lyme CSF/serum index; and further serology upon exposure. The test introduction was not accompanied with specific training of clinicians. Stewardship consisted of infectious diseases consults on request and general antibiotic treatment recommendation, but no formal stewardship chart review or audit in either cohort.
The primary study endpoint was the time to cessation of empiric antibiotic coverage against typical community-acquired bacterial meningitis pathogens (cephalosporins, carbapenems, vancomycin, and/or quinolones); secondary endpoints were the time to cessation of empiric coverage against Listeria monocytogenes (amoxicillin, trimethoprim-sulfamethoxazole, or aminoglycosides), as well as the time to cessation of antiviral coverage against HSV (acyclovir, valacyclovir, ganciclovir, or valganciclovir). Furthermore, length of hospital stay after lumbar puncture, as well as the proportion of pathogen-specific diagnoses achieved during respective hospital stays, were compared. Pathogen-specific treatment days for the respective positive diagnoses were not counted (eg, in case of HSV-1 encephalitis, targeted acyclovir treatment was not counted as unstopped empiric coverage).
Cox proportional hazards models were used to evaluate time-to-event endpoints, with adjustment for sex, age, immunosuppression, Charlson Comorbidity Index, CSF cell count, and C-reactive protein at CSF tap. Proportions were compared using χ2 or Fisher exact test, and continuous variables with t test or, where suitable, Kruskal-Wallis test. All analyses were performed using R software (R Foundation, Vienna, Austria).
RESULTS
In the PCR panel–based diagnostic pathway cohort, 198 patients were included, and in the culture-based diagnostic pathway, 393 patients were included. Characteristics of the 2 cohorts are given in Table 1.
Characteristics of Polymerase Chain Reaction Panel– and Culture-Based Cohorts
| Characteristic | PCR Panel | Culture Based | P Value |
|---|---|---|---|
| (n = 198) | (n = 393) | ||
| Age, y, median (IQR) | 56.83 (31.46–70.17) | 54.72 (39.50–67.40) | .68 |
| Sex | |||
| Male | 95 (48.0) | 184 (46.8) | .58 |
| Female | 103 (52.0) | 209 (53.2) | |
| CCI score | <.001 | ||
| 0 | 88 (46.8) | 90 (23.3) | |
| 1–2 | 80 (42.6) | 255 (66.1) | |
| 3–4 | 14 (7.4) | 37 (9.6) | |
| ≥5 | 6 (3.2) | 4 (1.0) | |
| Immunosuppressiona | 37 (18.7) | 55 (14.0) | .17 |
| CSF cell count, g/L, median (IQR) | 44.50 (14.25–127.75) | 85.00 (22.00–294.00) | <.001 |
| CSF glucose, mmol/L, median (IQR) | 3.33 (2.88–3.84) | 3.17 (2.21–3.99) | .05 |
| CSF protein, g/L, median (IQR) | 0.71 (0.44–1.05) | 0.78 (0.45–1.44) | .20 |
| CRP, mg/L, median (IQR) | 7.00 (1.00–26.00) | 23.50 (8.00–70.00) | <.001 |
| Pathogen-specific diagnosis | 30 (15.2) | 55 (14.0) | .81 |
| Streptococcus pneumoniae | 1 (0.5) | 3 (0.8) | 1 |
| Neisseria meningitidis | 0 (0.0) | 1 (0.3) | 1 |
| HSV-1, HSV-2, VZV | 9 (4.5) | 2 (0.5) | .002 |
| Listeria monocytogenes | 1 (0.5) | 2 (0.5) | 1 |
| Enterovirus | 6 (3.0) | 3 (0.8) | .08 |
| Tick-borne encephalitis | 4 (2.0) | 7 (1.8) | 1 |
| Lyme disease | 7 (3.5) | 8 (2.0) | .41 |
| Streptococci (non-pneumoniae) | 2 (1.0) | 1 (0.3) | .54 |
| Staphylococcus aureus | 0 (0.0) | 7 (1.8) | .14 |
| Gram-negative bacteria (non-Neisseria) | 2 (1.0) | 10 (2.5) | .35 |
| Other pathogen-specific | 2 (1.0) | 13 (3.3) | .16 |
| Characteristic | PCR Panel | Culture Based | P Value |
|---|---|---|---|
| (n = 198) | (n = 393) | ||
| Age, y, median (IQR) | 56.83 (31.46–70.17) | 54.72 (39.50–67.40) | .68 |
| Sex | |||
| Male | 95 (48.0) | 184 (46.8) | .58 |
| Female | 103 (52.0) | 209 (53.2) | |
| CCI score | <.001 | ||
| 0 | 88 (46.8) | 90 (23.3) | |
| 1–2 | 80 (42.6) | 255 (66.1) | |
| 3–4 | 14 (7.4) | 37 (9.6) | |
| ≥5 | 6 (3.2) | 4 (1.0) | |
| Immunosuppressiona | 37 (18.7) | 55 (14.0) | .17 |
| CSF cell count, g/L, median (IQR) | 44.50 (14.25–127.75) | 85.00 (22.00–294.00) | <.001 |
| CSF glucose, mmol/L, median (IQR) | 3.33 (2.88–3.84) | 3.17 (2.21–3.99) | .05 |
| CSF protein, g/L, median (IQR) | 0.71 (0.44–1.05) | 0.78 (0.45–1.44) | .20 |
| CRP, mg/L, median (IQR) | 7.00 (1.00–26.00) | 23.50 (8.00–70.00) | <.001 |
| Pathogen-specific diagnosis | 30 (15.2) | 55 (14.0) | .81 |
| Streptococcus pneumoniae | 1 (0.5) | 3 (0.8) | 1 |
| Neisseria meningitidis | 0 (0.0) | 1 (0.3) | 1 |
| HSV-1, HSV-2, VZV | 9 (4.5) | 2 (0.5) | .002 |
| Listeria monocytogenes | 1 (0.5) | 2 (0.5) | 1 |
| Enterovirus | 6 (3.0) | 3 (0.8) | .08 |
| Tick-borne encephalitis | 4 (2.0) | 7 (1.8) | 1 |
| Lyme disease | 7 (3.5) | 8 (2.0) | .41 |
| Streptococci (non-pneumoniae) | 2 (1.0) | 1 (0.3) | .54 |
| Staphylococcus aureus | 0 (0.0) | 7 (1.8) | .14 |
| Gram-negative bacteria (non-Neisseria) | 2 (1.0) | 10 (2.5) | .35 |
| Other pathogen-specific | 2 (1.0) | 13 (3.3) | .16 |
Data are presented as No. (%) unless otherwise indicated.
Abbreviations: CCI, Charlson Comorbidity Index; CRP, C-reactive protein; CSF, cerebrospinal fluid; HSV, herpes simplex virus; IQR, interquartile range; PCR, polymerase chain reaction; VZV, varicella zoster virus.
aSolid cancer with chemotherapy, hemato-oncological malignancy, human immunodeficiency virus, or immunosuppressive medication use.
Characteristics of Polymerase Chain Reaction Panel– and Culture-Based Cohorts
| Characteristic | PCR Panel | Culture Based | P Value |
|---|---|---|---|
| (n = 198) | (n = 393) | ||
| Age, y, median (IQR) | 56.83 (31.46–70.17) | 54.72 (39.50–67.40) | .68 |
| Sex | |||
| Male | 95 (48.0) | 184 (46.8) | .58 |
| Female | 103 (52.0) | 209 (53.2) | |
| CCI score | <.001 | ||
| 0 | 88 (46.8) | 90 (23.3) | |
| 1–2 | 80 (42.6) | 255 (66.1) | |
| 3–4 | 14 (7.4) | 37 (9.6) | |
| ≥5 | 6 (3.2) | 4 (1.0) | |
| Immunosuppressiona | 37 (18.7) | 55 (14.0) | .17 |
| CSF cell count, g/L, median (IQR) | 44.50 (14.25–127.75) | 85.00 (22.00–294.00) | <.001 |
| CSF glucose, mmol/L, median (IQR) | 3.33 (2.88–3.84) | 3.17 (2.21–3.99) | .05 |
| CSF protein, g/L, median (IQR) | 0.71 (0.44–1.05) | 0.78 (0.45–1.44) | .20 |
| CRP, mg/L, median (IQR) | 7.00 (1.00–26.00) | 23.50 (8.00–70.00) | <.001 |
| Pathogen-specific diagnosis | 30 (15.2) | 55 (14.0) | .81 |
| Streptococcus pneumoniae | 1 (0.5) | 3 (0.8) | 1 |
| Neisseria meningitidis | 0 (0.0) | 1 (0.3) | 1 |
| HSV-1, HSV-2, VZV | 9 (4.5) | 2 (0.5) | .002 |
| Listeria monocytogenes | 1 (0.5) | 2 (0.5) | 1 |
| Enterovirus | 6 (3.0) | 3 (0.8) | .08 |
| Tick-borne encephalitis | 4 (2.0) | 7 (1.8) | 1 |
| Lyme disease | 7 (3.5) | 8 (2.0) | .41 |
| Streptococci (non-pneumoniae) | 2 (1.0) | 1 (0.3) | .54 |
| Staphylococcus aureus | 0 (0.0) | 7 (1.8) | .14 |
| Gram-negative bacteria (non-Neisseria) | 2 (1.0) | 10 (2.5) | .35 |
| Other pathogen-specific | 2 (1.0) | 13 (3.3) | .16 |
| Characteristic | PCR Panel | Culture Based | P Value |
|---|---|---|---|
| (n = 198) | (n = 393) | ||
| Age, y, median (IQR) | 56.83 (31.46–70.17) | 54.72 (39.50–67.40) | .68 |
| Sex | |||
| Male | 95 (48.0) | 184 (46.8) | .58 |
| Female | 103 (52.0) | 209 (53.2) | |
| CCI score | <.001 | ||
| 0 | 88 (46.8) | 90 (23.3) | |
| 1–2 | 80 (42.6) | 255 (66.1) | |
| 3–4 | 14 (7.4) | 37 (9.6) | |
| ≥5 | 6 (3.2) | 4 (1.0) | |
| Immunosuppressiona | 37 (18.7) | 55 (14.0) | .17 |
| CSF cell count, g/L, median (IQR) | 44.50 (14.25–127.75) | 85.00 (22.00–294.00) | <.001 |
| CSF glucose, mmol/L, median (IQR) | 3.33 (2.88–3.84) | 3.17 (2.21–3.99) | .05 |
| CSF protein, g/L, median (IQR) | 0.71 (0.44–1.05) | 0.78 (0.45–1.44) | .20 |
| CRP, mg/L, median (IQR) | 7.00 (1.00–26.00) | 23.50 (8.00–70.00) | <.001 |
| Pathogen-specific diagnosis | 30 (15.2) | 55 (14.0) | .81 |
| Streptococcus pneumoniae | 1 (0.5) | 3 (0.8) | 1 |
| Neisseria meningitidis | 0 (0.0) | 1 (0.3) | 1 |
| HSV-1, HSV-2, VZV | 9 (4.5) | 2 (0.5) | .002 |
| Listeria monocytogenes | 1 (0.5) | 2 (0.5) | 1 |
| Enterovirus | 6 (3.0) | 3 (0.8) | .08 |
| Tick-borne encephalitis | 4 (2.0) | 7 (1.8) | 1 |
| Lyme disease | 7 (3.5) | 8 (2.0) | .41 |
| Streptococci (non-pneumoniae) | 2 (1.0) | 1 (0.3) | .54 |
| Staphylococcus aureus | 0 (0.0) | 7 (1.8) | .14 |
| Gram-negative bacteria (non-Neisseria) | 2 (1.0) | 10 (2.5) | .35 |
| Other pathogen-specific | 2 (1.0) | 13 (3.3) | .16 |
Data are presented as No. (%) unless otherwise indicated.
Abbreviations: CCI, Charlson Comorbidity Index; CRP, C-reactive protein; CSF, cerebrospinal fluid; HSV, herpes simplex virus; IQR, interquartile range; PCR, polymerase chain reaction; VZV, varicella zoster virus.
aSolid cancer with chemotherapy, hemato-oncological malignancy, human immunodeficiency virus, or immunosuppressive medication use.
The median time to cessation of empiric coverage against Streptococcus pneumoniae and Neisseria meningitidis was 3 days in the PCR-based pathway versus 4 days in the culture-based pathway (difference, −1 day [95% confidence interval {CI}, 0–3 days]; P = .16; Figure 1A).
Time to cessation (Cox proportional hazards model with 95% confidence intervals) of empirical antibiotic acute meningitis coverage (A), empirical Herpesviridae coverage (B), and Listeria monocytogenes coverage (C). Abbreviations: CSF, cerebrospinal fluid; HSV, herpes simplex virus; PCR, polymerase chain reaction.
The median time to cessation of empiric coverage against Herpesviridae was 1 day (95% CI, 0–1 day) in both cohorts, with no difference between the pathways (P = .93; Figure 1B).
No effect on time to cessation of empiric coverage against L monocytogenes was observed, with a median duration of 2 days in both pathways (difference, 0 days [95% CI, 0–3 days]; P = .69; Figure 1C).
The median length of hospital stay following lumbar puncture was not significantly shorter in the PCR panel–based pathway: 10 versus 12 days (difference, −2 days [95% CI, 0–3 days]; P = .86).
Last, the proportion of patients with a pathogen-specific meningitis or encephalitis diagnosis at discharge was not increased with the PCR panel–based diagnostic pathway (15.2% with pathogen-specific diagnosis vs 14.0% in the culture-based pathway; P = .80).
DISCUSSION
In this retrospective, single-center cohort study, we observed no significant shortening of empiric antibiotic coverage against community-acquired meningitis–causing pathogens and length of hospital stay with the introduction of a PCR panel–based CSF diagnostic pathway (BioFire FilmArray Meningitis/Encephalitis Panel). Time to stop of antimicrobial coverage against Herpesviridae and L monocytogenes remained unchanged with the added CSF-PCR panel.
The added value of PCR panel–based CSF diagnostics through shorter test turnaround times is well established; however, the proportion of pathogen-specific diagnoses did not increase with the test panel in our cohort. Upon analysis of the specific meningitis or encephalitis pathogens, the typical community-acquired meningitis–causing bacteria were detected to a similar degree in the culture-based pathway cohort.
The additional benefit on reduced use of antimicrobials to date has only been evaluated in small cohorts with conflicting unadjusted results. We found a non significant shortening of empiric antibiotic coverage duration against acute bacterial meningitis causing pathogens. Of note, median empiric coverage in the PCR-based pathway was still substantially longer than the test turnaround time, potentially due to ongoing ceftriaxone application until serological exclusion of neuroborreliosis in some cases.
Despite the excellent negative predictive value of the test panel for L monocytogenes [6], the introduction of the test did not lead to a reduction of the corresponding empirical coverage time, despite the large difference in test turnaround time (3 hours for the PCR panel vs 5 days in CSF culture). The reasons for this are unclear, but we suspect they include factors such as lower confidence in the test and its ability to exclude this specific diagnosis compared to conventional CSF culture among treating physicians. This could potentially be improved by adequate education on test characteristics at introduction.
We hypothesize that the lack of improvement in empiric antiviral coverage duration against Herpesviridae may be due to the long-standing availability of a dedicated PCR test with a comparable test turnaround time (4 vs 3 hours at our institution), which preceded the introduction of the PCR panel studied here.
The introduction of a PCR panel–based diagnostic pathway failed to significantly shorten length of hospital stay in our cohort. In context, earlier studies reported conflicting results regarding the influence on length of stay with 1 study demonstrating a reduction in duration of hospitalization [7] and 1 study not demonstrating this reduction [5].
Our study has several limitations. It is a retrospective single-center analysis on the effects of 2 different test pathways. The 2 cohorts may have differed by being apart from 1 to several years; therefore, evolving practices could have influenced outcomes. Some selection bias cannot be excluded (which is also true for earlier studies on this subject), as some meningitis cases in the time after introduction of the PCR panel test may have been evaluated without the use of the PCR panel.
CONCLUSIONS
PCR panel–based diagnostic pathways for meningitis/encephalitis evaluation are a very helpful tool in clinical practice, given the rapid test turnaround time. However, their effect on use of antimicrobials and duration of hospital stay might be overestimated: We found no significant reduction in the duration of empiric antimicrobial coverage against acute bacterial meningitis pathogens and length of hospital stay, and observed no effect regarding empirical antimicrobial treatment against Herpesviridae or L monocytogenes. This findings should be validated in future larger multicenter studies.
Factors such as lacking trust of prescribers in negative test results in a newly introduced test as compared to the historical gold standard of culture (along with a preexisting, dedicated HSV-PCR) may contribute to suboptimal discontinuation of empiric treatments. Further research should explore factors influencing the motivation to discontinue empirical coverage to fully exhaust the benefits conferred by tests with improved test turnaround time. Test introductions should be accompanied with adequate education of clinicians on test characteristics.
Notes
Patient consent. The study was approved by the local ethics committee (Cantonal ethics committee Bern, Switzerland, approval B2019-01902) and used data of patients with signed general consent for the further use of data for research. Patients with objection for further use of their clinical data for research were excluded.
Financial support. No external funding, the study was covered by department funds.
Potential conflicts of interest. The authors: No reported conflicts of interest.
Comments