Background
Meningitis remains a major cause of mortality and morbidity worldwide. In Ethiopia, it is an important cause of premature death and disability, being the 9th most common cause of years of life lost and loss of disability-adjusted life years [
1]. In addition to the morbidity and mortality associated with meningitis, the disease also represents a huge burden on affected families and the health care system, especially in low-income countries. One case of meningococcal disease in sub-Saharan Africa has been estimated to cost the household 90 US dollar (USD), equal to 34% of the annual gross domestic product per capita, and an additional 154 USD if the disease leads to sequelae [
2].
Early treatment is essential in the clinical management of meningitis. A delay in therapy negatively affects the prognosis for patients with both bacterial [
3] and viral [
4] meningitis/encephalitis. Empiric antimicrobial therapy is therefore initiated on suspicion, before laboratory results are available [
5,
6]. Traditionally, laboratory diagnostics of meningitis have included examination of cerebrospinal fluid (CSF) for white blood cells (WBC), measurement of glucose and protein levels, and Gram staining. These are rapid, low-cost analyses but do not have satisfactory specificity alone [
7,
8]. When septicemia is additionally suspected, peripheral blood WBC count is usually obtained, but this analysis also suffers from a lack of sensitivity and specificity [
9,
10]. Positive bacterial cultures from CSF can provide a definite diagnosis, but may take several days and quality requirements are high. Therefore, rapid molecular diagnostics have become routine in meningitis diagnostics in high-income countries.
In many low-income countries, poor laboratory services impede the quality of health care. Ethiopia is experiencing high gross domestic product growth, but the per capita income remains one of the lowest in the world [
11]. Healthcare in Ethiopia is under-resourced, and laboratory services are weak. Inadequate and delayed microbiological services foster distrust in the value of laboratory testing among clinicians, and reduce their inclination to submit clinical samples for analyses. Ultimately, this compromises patient care, prevention of infectious diseases and antibiotic stewardship.
More often than not, in Ethiopia and other low-resource settings patients receive an inadequate diagnostic workup [
12] and decisions to treat meningitis are based on clinical findings alone. This approach lacks accuracy [
13,
14] and ultimately results in unjustified prescriptions of antibiotics [
12]. Incorrect use of antimicrobials is a recognized challenge in Ethiopia, with delayed initiation, unindicated use, duplication of broad spectrum antimicrobials and unnecessarily long duration of treatment being the major problems [
15]. Fortunately, in the last few years greater attention has been paid to antimicrobial resistance also in Ethiopia, where it is becoming an increasing concern [
16].
Recently, a rapid molecular diagnostic system (FilmArray, BioFire;
http://www.biomerieux-usa.com/clinical/biofire-film-array) was developed for multiplex testing for the most common causes of infectious syndromes. The system is closed and integrates nucleic acid extraction and nested PCR for detection of the pathogens included in the panels. It is easy to use, has a low risk of sample contamination, requires little hands-on time and limited prior training in molecular diagnostics. For each sample, a kit that contains all necessary reagents is used, and sample preparation takes only a couple of minutes. These features are advantageous in settings with limited experience and limited facilities for molecular analyses. The main limitation for use in such settings is the high cost of test-kits.
The aim of this study was to investigate which infectious agents were responsible for meningitis in a region of Ethiopia where treatment decisions are usually made without microbiological results. We demonstrated that an easy-to-use and reliable molecular diagnostic instrument could set the ground for improving patient management and reducing usage of antimicrobials in a resource-limited setting.
Methods
Study design and population
This prospective study was performed between March 20th and June 20th 2017, at Jimma University Specialized Hospital (JUSH), one of the oldest public hospitals in Ethiopia, established in 1937. It is the only teaching and referral hospital in the southwestern part of the country and serves a catchment population of 15–20 million. The hospital has 600 beds and 1621 staff members, providing services for 218,095 outpatients, 16,778 inpatients, 14,207 emergencies and 5,973 deliveries yearly (data from September 2016 to September 2017). Other public health services in the catchment area include five primary hospitals, one army hospital, 118 health centres and 482 health posts.
All patients with a clinical suspicion of meningitis and from whom a CSF sample was available for analyses were eligible for inclusion. The study was run as an integrated part of routine practice and inclusion was done by the treating physicians. A written consent was given by the participants, or by parents or guardians in the case of children or unconscious patients. As samples were taken as part of routine diagnostic practice, participation in the study required no additional interventions.
Before the start of the study, a general introduction of the project aims and procedures, to the new instrument and interpretation of results was given to all medical doctors at the pediatric and internal medicine department. A four-hour training on how to perform analysis with the FilmArray was given to four dedicated JUSH staff members at the bacteriology department. Results of the PCR analysis are generated automatically by the FilmArray software and are reported as either negative, or positive with a defined species. Two senior medical doctors (study investigators) associated with the pediatric and medical wards, respectively, acted as supervisors for other physicians on the wards during the project. They were responsible for follow-up, and supported treating physicians with clinical/treatment implications of the laboratory results. The first author, a medical doctor, was present at JUSH during the project period and followed up data collection. Sample analyses, reporting of results and clinical decisions about treatment were done by local JUSH staff.
Clinical data, treatment, routine investigations and outcome
Patient data were collected on a structured case record form. The treating physician collected the consent and clinical data at admission while microbiological analysis and results were documented by laboratory staff. Laboratory results were, according to routine at JUSH bacteriology laboratory, collected by the treating physician and the paper transcript of the laboratory result was kept in the patient chart. Results were also intended to be immediately reported to the treating physician by cellular phone. However, the latter additional reporting was not always complied with in practice. Data on treatment, routine investigations and outcome were collected by study investigators (all medical doctors), either at discharge or retrospectively within one month of discharge. Patients with none or incomplete data on treatment, routine investigations and outcome were excluded from the relevant statistical/data analyses, but included in analyses on clinical and laboratory data.
Sample collection
CSF samples were collected in sterile glass tubes for neonates and paediatrics and in sterile plastic tubes (5 ml, Sartstedt, Germany) for adults. They were transported to the laboratory at ambient temperature within ten minutes and laboratory analyses were started within two hours, but usually within 30 min. For patients with suspected concomitant septicemia, blood was collected in blood culture bottles using BD BACTEC Peds Plus (Becton Dickinson (BD), Franklin Lakes, NJ, USA) for paediatric patients and BD BACTEC Plus Aerobic medium (BD) for adult patients.
Conventional laboratory analyses of CSF
CSF samples were processed according to standard procedures at JUSH which include bacteriological culture, macroscopic assessment, protein and glucose measurements, manual WBC count with differential count and Gram staining. India ink staining and rapid cryptococcal antigen testing (CrAg, IMMY, OK, USA) were performed upon request. Gram stain was performed on the primary sample, usually without centrifugation due to lack of equipment and/or consumables. CSF glucose and protein measurements, were performed when reagents were available. CSF samples were cultured onto blood agar (BD) and/or chocolate agar (BD) for 72 h at 35 °C in a CO
2-enriched incubator (candle jar). Routine bacterial identification was based on colony morphology, Gram staining and standard biochemical reactions [
17].
Molecular diagnostic of CSF
All CSF samples included in the study were analysed using the meningitis/encephalitis (ME) panel (bioMérieux, Marcy l’Etoile, France) on the FilmArray (bioMérieux,) multiplex PCR system. The closed system performs nucleic acid extraction, reverse transcription and multiplex nested PCR, automatically interprets end-point melting curve data and provides the result. The ME panel tests for genetic targets from
Escherichia coli K1,
Haemophilus influenzae,
Listeria monocytogenes,
Neisseria meningitidis,
Streptococcus agalactiae,
Streptococcus pneumoniae, cytomegalovirus, enterovirus, herpes simplex virus 1 (HSV-1), herpes simplex virus 2 (HSV-2), human herpesvirus 6 (HHV-6), human parechovirus, varicella zoster virus (VZV) and
Cryptococcus neoformans/gattii. Analyses were performed according to the manufacturer’s instructions [
18]. A flow-chart of the sample analysis with the FilmArray is shown in Additional file
1: Figure S1.
Other laboratory analysis
CSF samples from patients with suspected tuberculous meningitis (clinical suspicion, HIV-positivity, or no improvement on conventional antibiotic treatment) were tested for
Mycobacterium tuberculosis with the GeneXpert MTB/RIF assay (Dx System Version 4.0c, Cepheid Inc., CA, US) [
19]. CSF was processed as recommended by the manufacturer: a 2:1 volume of the reagent supplied with the assay was added directly to the sample. The mixture was vortexed and incubated at room temperature for 15 min. Two mL of the reagent-sample mix was then transferred to the Xpert cartridge and analysed.
Blood culture bottles were incubated in an automated blood culture incubator BACTEC FX40 system (BD) for a maximum of five days and flagged as either “negative” or “positive” for bacterial growth. Positive blood cultures were analysed using the FilmArray blood culture identification (BCID) panel (BioMerieux, France) [
20], and cultured onto blood agar (BD), MacConkey agar (BD) and chocolate agar (BD), with the same routine bacterial identification as for CSF cultures.
Other investigations, such as malaria testing using peripheral blood smear and May-Grünwald-Giemsa staining, rapid testing for HIV, haematological and biochemical analyses were performed according to routine practice at JUSH.
Data management and statistical analysis
All data were entered into a Microsoft Excel 2013 database. Statistical analyses were performed in Microsoft Excel 2013 or in GraphPad Prism 5. Two-sided tests were used and a p-value < 0.05 was considered statistically significant.
Discussion
We analysed clinical and microbiological data from prospectively recruited patients with suspected meningitis in a teaching hospital in Ethiopia. The study was performed within the routine clinical and laboratory settings of a hospital that had very limited prior experience with molecular techniques. In addition to the conventional laboratory investigations used in the hospital, a simple and rapid molecular diagnostic system was introduced to enhance laboratory diagnostics during the study period. To the authors’ knowledge, this is the first time a definite etiological diagnosis of viral meningitis has been made in patients in a public health facility in Ethiopia.
Virus were the most common etiological agents of meningitis in this study with HHV-6 being the most common, followed by HSV-1 and HSV-2 and enterovirus. Although the etiology is likely to vary between age groups and local epidemiology, this is in agreement with a study from Finland that found enteroviruses, followed by HSV-2 and VZV to be major causes of aseptic meningitis in adults [
21] and a study from Brazil that reported enterovirus as the most common cause of meningitis, followed by HSV-1, cytomegalovirus and dengue virus [
22].
Early detection of etiologic agents improves the outcome of meningitis [
23] and adequate laboratory diagnostics are imperative. Culturing of bacterial and fungal agents takes time and has low sensitivity, as illustrated by the fact that only one of 9 CSF samples with potentially cultivatable organisms was culture positive. The sensitivity of culture is affected by many factors including prior administration of antibiotics, suboptimal culturing conditions and media, and fastidious nature of some of the bacterial agents. Because of this and the unavailability of viral detection, virtually all patients with suspected meningitis in Ethiopia are treated as bacterial meningitis cases and the diagnosis is rarely re-evaluated over the course of the disease. This over-diagnosis of bacterial meningitis inevitably leads to an overuse of antibiotics. Hence, rapid molecular diagnostics can have a major impact in low-income settings by increasing the likelihood of reaching a correct diagnosis and enabling correct patient management. This is crucial not only for the outcome of the individual patient, but also for hospital biosecurity measures, public health decisions and both local and global efforts to reduce and improve antimicrobial usage.
The introduction of syndromic testing of infectious diseases and fully automated multiplexed analyses represents a paradigm shift in microbiological diagnostics [
24]. Compared to the conventional laboratory investigations performed at JUSH, that detected a microorganism in CSF of only five patients (two by Gram stain, one by Gram stain and cryptococcal antigen-testing, one by cryptococcal antigen only and one by culture), the FilmArray was able to detect microorganisms in 20 samples using the ME panel. Hence, such systems can improve patient management in settings with limited laboratory facilities. The FilmArray system was very easily implemented into a modestly equipped laboratory where personnel had little prior experience with molecular diagnostics. The laboratory personnel were able to take on the extra task without disrupting their regular working schedule, while the machine operates.
However, there are a number of limitations to the sustained use of such automated systems in low-income countries. In Ethiopia, procurement of the necessary consumables is a complicated and lengthy process. The main obstacle, however, is the cost. Currently, the reagents needed for the analysis of one sample exceed 100 USD. Needless to say this is not sustainable in a public health system that is already financially constrained. On the other hand, a full course of treatment for suspected bacterial meningitis for 10–14 days [
25], including only direct expenses for a hospital stay, routine investigations and antibiotic treatment, is likely to amount to more than 100 USD, even in Ethiopia. Another possible limitation of the system is the predefined selection of the pathogens in the panels. The panels were developed for an American market and may not be equally suited for Africa where other pathogens including
M. tuberculosis and malaria are major causes of infections. It is also important that clinicians have a good understanding of test characteristics, interpretation of results and test limitations. Although lower than for conventional PCR, there is still a potential for sample contamination when using the FilmArray and the assay may detect latent or reactivated viruses [
26]. The assays should be used with care and the positivity rates should be monitored.
Molecular diagnostics should not replace conventional methods. Simple investigations like WBC, Gram stain, and glucose and protein measurements provide valuable diagnostic information, helping distinguish between bacterial and non-bacterial meningitis and guiding treatment [
7,
23,
27‐
30]. These simple methods, however, which are defined as routine investigations at JUSH, were not available for most of the study period. No patients in this study had blood/CSF glucose ratio measured, mainly because of challenges in procurement and finances leading to a lack of basic reagents. Nevertheless, the study did confirm that manual cell counts were useful and could be used more actively to influence patient management.
A combination of laboratory diagnostics and clinical examination have been shown to provide the best prediction of bacterial meningitis, with 100% sensitivity and 52% specificity [
28], but even CSF cell count alone can be a good predictor. A cut-off value of 321 leukocytes/mm
3 showed a sensitivity and specificity for bacterial meningitis of 81% in a paediatric population in Portugal [
27] and all patients with bacterial meningitis in a mixed-age population in Egypt were found to have > 100 leukocytes/mm
3 [
29]. In our study, all confirmed cases of bacterial meningitis had ≥75 leukocytes/mm
3. Thus the use of a cut-off value of ≥75 leukocytes/mm
3 and/or a turbid appearance of CSF to indicate bacterial meningitis and the need for antibiotic treatment would have reduced the number of patients treated with antibiotics by almost 75% in this study.
The high number of patients with no cells or microorganisms detected in the CSF samples indicates that meningitis may be overestimated, especially in children and neonates. Detection of microorganisms in CSF of only 4% of neonates, might suggest that lumbar puncture, an invasive procedure, is overused. In contrast, the high proportion of positive blood cultures in this study (38%), particularly in neonates (67%) suggests that this diagnostic procedure is underused in the hospital.
A limitation of the study is that, despite close follow-up, medical files with treatment and outcome data were not traceable for a considerable proportion (27%) of the patients. This may have introduced a bias with regards to evaluating outcome. However, the proportion of patients where a microorganism was detected in CSF was the same for the overall study population (9.9%) as for the population where outcome was recorded (9.6%).
No differences were observed in this study with respect to prescription of antibiotics to patients with bacterial/fungal agents detected in CSF versus those with viral meningitis or no microorganism detected. This could be due to a lack of clinical treatment guidelines, but a contributing factor might have been poor communication of test results to responsible clinicians due to lack of electronic reporting systems, operational hospital telephones and internal post delivery systems. We acknowledge that changes in prescription habits will require long lasting interventions and coordinated stewardships in order to be successful, and that the short duration of the study was a limitation, in this respect. A recent publication from India observed a target specific escalation and evidence based de-escalation of the use of antimicrobials when a syndromic based molecular diagnostic system for meningitis was used over a period of 4 years [
31].
The prescription of antibiotics is far higher in Ethiopia than WHO recommendations [
32]. There is no formal antimicrobial stewardship program at JUSH, nor restrictions or specific guidelines on antimicrobials use. In Ethiopia, there is no national policy or guidelines on antimicrobials use, neither do most hospitals have their own systems to ensure correct use of antimicrobials [
15]. A study from JUSH also showed that the national guidelines did not cover the management of 20% of diagnosed infectious diseases [
15]. Ensuring stable access to simple CSF investigations, including protein and glucose measurements, and developing and implementing treatment guidelines for meningitis based on laboratory results might be a first and important step to improve treatment and reduce the misuse of antibiotics.