Evaluation of the FilmArray Meningitis/Encephalitis Panel for Diagnosis of Infectious Meningitis and Encephalitis

Background: Infectious Meningitis/Encephalitis (M/E) is caused by pathogens. The FilmArray M/E panel can quickly detect 14 kinds of pathogens and facilitate diagnosis for patients with M/E. This study aimed to perform a evaluation of the sensitivity and specicity of the FilmArray M/E panel compared with classic method of diagnosing M/E. Methods: Relevant studies published before August 15, 2021, were identied by searching Web of Science, PubMed, Cochrane Library, and Embase, using keywords Meningitis, Encephalitis, and FilmArray M/E panel. After the initial screening, EndNoteX9 was used to manage the studies and extract data. The quality of the study was based mainly on the quality evaluation standard of diagnostic tests recommended by Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2). Meta-DiSc 1.4 statistical software was used to determine the sensitivity, specicity, 95% condence interval (CI), diagnostic odds ratio (DOR), positive likelihood ratios (PLR) and negative likelihood ratios (NLR) of each group of data, and summary receiver operating characteristic curve (SROC). All P values were two sided, and P <0.05 was considered statistically signicant. Results: A total of 269 studies were retrieved, and 16 full-text studies were identied. The sensitivity of all the full-text studies was 0.93 (95% CI: 0.91–0.95) and the specicity 0.98 (95% CI: 0.98–0.98). Conclusions: Compared with the gold standard, the FilmArray M/E panel had a sensitivity of 0.93 and a specicity of 0.98 to 16 pathogens. Further studies are needed to determine whether the FilmArray M/E panel can be used as a clinical standard for the diagnosis of M/E.


Introduction
Generally speaking, infectious meningitis and encephalitis almost have no difference 1 . Meningitis/encephalitis (M/E) is caused by infection by pathogens such as viruses and bacteria. These pathogens may also be fungi, parasites, and so on 2 . Patients with meningitis/encephalitis have high mortality and morbidity. M/E is prone to some serious sequelae such as epilepsy, audio-visual disorder, limb dyskinesia, and so forth, and may even be life-threatening 3,4 . Primary bacterial pathogens are responsible for more than 80% of community-acquired acute bacterial meningitis, such as Haemophilus in uenzae, Neisseria meningitidis, Listeria monocytogenes, Lance eld Group B streptococci, and Streptococcus pneumoniae 5 . However, meningitis caused by the virus is usually benign with no sequelae in patients with normal immunity 1 . It has been reported that the main pathogens detected in patients with viral M/E are enterovirus and Human herpesvirus 6 6 . Because of the different types of pathogens, the treatment of meningitis and encephalitis is quite different 7 . In the past, the method of diagnosing M/E was ine cient and the accuracy was not high 2 . Hence, a more complete method for diagnosing M/E was urgently needed 8 .
The FilmArray M/E plate is a multi-molecular panel approved by the FDA in 2015 for the rapid detection of 14

Inclusion and exclusion criteria
The inclusion criteria were as follows: (1) clinical samples were tested using the M/E panel; (2) other commonly used diagnostic methods were employed to analyze the samples and compared with the results of the FilmArray M/E panel; and (3) the 2 × 2 table was extracted based on the experimental results, and the speci city and sensitivity were calculated.

Data extraction
EndNoteX9 was used to manage the studies and extract data after the preliminary screening. Data extracted included country, author, literature year, sample source, study design type, sample type, gold standard, true positive, false positive, true negative, and false negative.

Quality assessment
The methodological quality of each study was assessed using the validated Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool 13 . Two investigators independently evaluated the risk of bias in each study according to the QUADAS. Total risk of bias in every study was assessed with "Yes," "Unclear," and "No." Discrepancies were resolved by discussion to get a consensus assessment. In the case of persistent disagreement, an adjudicator was consulted.

Statistical analysis
The Meta-DiSc 1.4 software was used to analyze the data from reconstructed 2 × 2 tables: sensitivity, speci city, negative likelihood ratio (NLR), positive likelihood ratio (PLR), 95% con dence interval (95% CI), and diagnostic odds ratio (DOR). A bivariate mixed model was adjusted to obtain a summary receiver operating characteristic (SROC) curve by summarizing the joint distribution of sensitivity and speci city with the Moses linear model, and the corresponding area under the curve (AUC) was calculated as a global measurement of test performance 14,15 . To avoid empty cells in 2 × 2 tables, "0.5" was used to replace the cell "0". Statistical tests were generally two sided, and therefore only P values < 0.05 were considered statistically signi cant.

Search results
A total of 269 studies were retrieved and 129 studies were excluded due to duplication. Of the remaining 140 studies, 31 were excluded after browsing the titles and abstracts and 92 (66 insu cient information, 13 letter, 1 Meta-analysis and systematic review, 4 review and 8 case reports) were excluded after screening the full-text.
Eventually, 16 studies and in total 17 datasets were selected for full-text review and evaluation 2-4,16−25 . The elaborated process of this literature search is presented in S1 Fig.

Study characteristics
Among the 16 studies till July 15, 2021, 15057 samples with no age restriction were included. The "total" in Table 1 represents the overall datum (including bacterial, viruses, and fungi) of each study. Bacteria, viruses, and fungi were used as subgroups to extract data for each article. Eventually, 16 studies had bacterial data, 10 had viruses' data, and 7 had fungi data. In addition, a study could only extract the total data. The characteristics of these studies are presented in Table 1.

Risk-of-bias assessment
A summary of the risk of bias assessment results are shown in Table 2. Each of the 11 components according to QUADAS-2 criteria was graded as "yes", "unclear" or "no", which meant "low risk of bias", "uncertain of bias" and "high risk of bias" respectively, based on the methods reported in each study.

Diagnostic accuracy of FilmArray for infectious meningitis and encephalitis
Pooled sensitivity and speci city of FilmArray in meningitis and encephalitis are shown in Fig. 1(A and B).
Diagnostic accuracy of FilmArray in infectious meningitis and encephalitis for bacteria Among the 16 included studies, the pooled sensitivity (0.92, 95% CI 0.87-0.96) and speci city (0.99, 95%CI 0.99-0.99) in diagnosing meningitis and encephalitis for bacteria are displayed in Fig. 5(A and B).

Diagnostic accuracy of FilmArray in meningitis and encephalitis for viruses
Among the 10 included studies, the pooled sensitivity (0.96, 95% CI 0.94-0.98) and speci city (0.99, 95%CI 0.99-1.00) in diagnosing meningitis and encephalitis for viruses are displayed in Fig. 6(A and B).
Diagnostic accuracy of FilmArray in infectious meningitis and encephalitis for fungi Among the 7 included studies, the pooled sensitivity (0.91, 95% CI 0.59-1.00) and speci city (1.00, 95%CI 1.00-1.00) in diagnosing meningitis and encephalitis for fungi are displayed in Fig. 7(A and B).

Discussion
The use of the FilmArray M/E panel started in October 2015. The 14 pathogens detected in CSF included 6 bacteria, 7 viruses, and 1 yeast 9 . In the present study, 269 published studies were searched and 16 were extracted after screening. The data were analyzed to evaluate the accuracy of the FilmArray M/E panel. The whole data displayed that the sensitivity was 0.93; speci city was 0.98; PLR was 20.98; NLR was 0.09; DOR was 286.21; and SROC AUC was 0.9765. The FilmArray M/E panel had sensitivity to bacteria, viruses, and fungi of 0.92, 0.96, and 0.91, with speci city of 0.99, 0.99, and 1.00, respectively. PLR was greater than 10 and NLR was 0.13, which was close to 0.1. The DOR indicated that the probability of correct diagnosis was 148.89 times that of diagnostic errors. The AUC in the SROC curve was closer to 1, and the SROC curve was closer to the upper left corner. These ndings indicated that the overall diagnostic accuracy of the FilmArray M/E panel was high. In addition, viruses, bacteria, and fungi had high sensitivity and speci city. The data on fungi was insu cient; hence, the speci city was shown as 1. In addition, the coe cient of bias P = 0.391 > 0.05, indicating that the probability for publication bias was subtle.
A total of 16 studies were retrieved, and 15057 samples were detected using the FilmArray M/E panel. The de nite diagnosis of bacterial meningitis has historically been based on culture, with a sensitivity of ≤ 80%. Another study can be used for comparison, which is "Evaluation of the BioFire FilmArray M/E panel for the detection of bacteria and yeast in Chinese children 18 ." A total of 223 patients were detected, and 68 CSF samples met the inclusion criteria. The mean age of the patients was 2.76 years (from 3 days to 12 years), and the male-female ratio was 2.09 (46:22). The present study focused mainly on younger patients. In this paper, the ability of the FilmArray M/E panel in the diagnosis of bacterial and fungal meningitis has been demonstrated through the study of children in China. However, in our paper, we analyzed the data of virus, bacteria and fungi in detail and comprehensively, and analyzed the situation of patients of all ages. This meta-analysis had several limitations. First, no virus comparison results were used. Second, the sample size was small, affecting the statistical certainty of the sensitivity and speci city calculations of the FilmArray M/E panel. The other study entitled "Diagnostic test accuracy of the BioFire® FilmArray® ME panel: a systematic review and meta-analysis," screened 3059 patients. The gold standard for this study was consistent with that used in the present analysis. It concentrated on the analysis of speci city and sensitivity. Moreover, its sensitivity and speci city were both > 90%. The high accuracy estimates pointed out that the FilmArray M/E panel could be a very useful adjunct tool for the diagnosis of meningitis/encephalitis. This clinical evaluation had a few restrictions. First, only 13 studies met the inclusion criteria, leading to a small sample size and limiting the extrapolation of the results. In addition, despite collecting all relevant studies, it was still di cult to ensure that no data were missed. Furthermore, the use of different diagnostic criteria led to the heterogeneity among the selected studies. This analysis can be further improved by accumulating clinical data in the future.
In summary, the FilmArray M/E panel is a method with high sensitivity and speci city when used for the diagnosis of meningitis and encephalitis. It is a rapid, sensitive, and speci c detection method valuable for the clinical diagnosis of meningitis and encephalitis.

Conclusion
The FilmArray M/E panel is a tool with high diagnostic performance. However, more research is needed to explore its effects.    28 Lumley 26 Lee 23 Du 19 Messacar 27 Graf 20        Deeks' funnel plot asymmetry test to assess publication bias in estimates of diagnostic odds ratio for FilmArray M/E panel detection of meningitis and encephalitis.