Mouse models of Japanese encephalitis virus infection: A systematic review and meta-analysis using a meta-regression approach

Background Japanese encephalitis (JE) virus (JEV) remains a leading cause of neurological infection across Asia. The high lethality of disease and absence of effective therapies mean that standardised animal models will be crucial in developing therapeutics. However, published mouse models are heterogeneous. We performed a systematic review, meta-analysis and meta-regression of published JEV mouse experiments to investigate the variation in model parameters, assess homogeneity and test the relationship of key variables against mortality. Methodology/ Principal findings A PubMed search was performed up to August 2020. 1991 publications were identified, of which 127 met inclusion criteria, with data for 5026 individual mice across 487 experimental groups. Quality assessment was performed using a modified CAMARADES criteria and demonstrated incomplete reporting with a median quality score of 10/17. The pooled estimate of mortality in mice after JEV challenge was 64.7% (95% confidence interval 60.9 to 68.3) with substantial heterogeneity between experimental groups (I^2 70.1%, df 486). Using meta-regression to identify key moderators, a refined dataset was used to model outcome dependent on five variables: mouse age, mouse strain, virus strain, virus dose (in log10PFU) and route of inoculation. The final model reduced the heterogeneity substantially (I^2 38.9, df 265), explaining 54% of the variability. Conclusion/ Significance This is the first systematic review of mouse models of JEV infection. Better adherence to CAMARADES guidelines may reduce bias and variability of reporting. In particular, sample size calculations were notably absent. We report that mouse age, mouse strain, virus strain, virus dose and route of inoculation account for much, though not all, of the variation in mortality. This dataset is available for researchers to access and use as a guideline for JEV mouse experiments.


Unfunded studies
Enter: The author(s) received no specific funding for this work. Describe where the data may be found in full sentences. If you are copying our sample text, replace any instances of XXX with the appropriate details.
If the data are held or will be held in a public repository, include URLs, accession numbers or DOIs. If this information will only be available after acceptance, indicate this by ticking the box below. For example: All XXX files are available from the XXX database (accession number(s) XXX, XXX.

meta-analysis using a meta-regression approach
We would be grateful if you could consider our submission describing a systematic review and meta-analysis of mice models of Japanese encephalitis virus.
Japanese encephalitis (JE) is a devastating disease, with no promising antiviral candidate therapies. Treatment trials that have been conducted were only powered to detect very large treatment effects. There are still around 100,000 cases per year despite the availability of vaccines. In order to have any hope of developing new treatments, standardised animal models will be needed. However, mouse models of JE are variable in their characteristics, with many contradictory findings in the literature. For this reason, we performed a systematic review, meta-analysis and meta-regression of published JEV mouse experiments, to investigate the variation in model parameters, assess homogeneity, and test the relationship of key variables against mortality. This is the first report, we are aware of, of a systematic review and meta-analysis of mouse models of JE. We demonstrate an abundance of experimental work in this field. However, reporting was frequently incomplete and there was considerable variability in outcomes. We provide recommendations for researchers to improve standardisation of study design and reporting of studies. Furthermore, all the data have been provided in supporting files.  The high lethality of JE and absence of effective therapies mean that animal models are 120 crucial in developing our understanding of the disease and therapeutic prospects. There 121 have been reports published of experiments using JEV in mouse models dating back to 122 1935 when the virus was first isolated (9, 10). Mice are frequently the preferred model for 123 studying human infections due to their low-cost, timely reproduction and variability (11). 124 In the last decade, many studies of JEV have used animal models to address a wide 125 variety of different questions, such as the role of various components of the immune 126 system in protection from JEV, pathogenesis of JE and for testing treatments. These 127 studies have contributed greatly towards further understanding JE pathogenesis (7). 128 However, mouse models of JE have not been standardised and can be highly variable 129 across laboratories contributing to contradictory results (12). The study adhered to PRISMA guidelines for systematic reviews and the protocol is 141 included in supplementary data (S1_Data). A PubMed search was performed using the The inclusion criteria were any publication that included an experiment meeting the 152 following criteria: 1. JEV was inoculated into mice; 2. virus dose was reported; 3. JEV 153 strain or source was reported; 4. immunocompetent mice were used and the strain was 154 reported; 5. mortality was reported as either death or humane endpoint (primary 155 analysis & outcome measures, secondary outcome measures 1 and 2) or other 156 pathological outcome reported (secondary outcome 3); 6. published in English; and 7. 157 primary research. Additional data were also extracted (see S1_Data) but did not serve 158 as an exclusion from the primary analysis. Studies were excluded if they reported 159 inoculation using non-pathogenic JEV (for example the vaccine strain SA14-14-2) only, 160 or if data on individual animals was not reported and it was not possible to extract the 161 data. In order to minimise non-specific immune effects, data were collected only on 162 groups of mice that received JEV only, and no other material. For example, data were 163 frequently derived from reports that tested a treatment or vaccine, in these cases only 164 the control group was extracted. 165 Quality assessment and data extraction were performed by two authors independently, were not reported. Less than 50% of studies included statements that allocation to 209 experimental group was random, treatment and outcome were blinded to investigator, 210 13 neuroprotective anaesthesia was used and no conflict of interests were present. No study 211 reported having performed a sample size calculation or that the temperature was 212 controlled during experiments. Median quality score was 10 (IQR 9-11, range 7-13) 213 across the 17 criteria (figure 2). Data extracted are displayed in Table 1  Mouse strain: Fourteen different mouse strains were used in the studies, see Figure 3A. Quality assessment was relatively low with a median quality score of 10 (IQR 9-11, range 373 7-13) across the 17 criteria. However, there was an improvement in the score over time. 374 This highlights the need to reduce risk of bias through detailed reporting, following 375 CAMARADES guidelines. In particular, there is a need for performance and reporting of 376 blinding, randomisation and a priori sample size calculations. None of the studies were 377 excluded based on these quality assessments. Beyond the quality assessment criteria, 378 analysis of data extracted demonstrated missing data for study-level characteristics, most 379 strikingly for reporting of mouse sex in 59 (56%) studies. Furthermore, accession 380 numbers were only reported in 20 (16%) studies with no study reporting sequencing of 381 21 the actual JEV strain used, which is important given the potential for sequence variability 382 after serial passage in culture. 383 The number of mice per experimental group has become more consistent with time, as 384 researchers are more aware of the need to reduce unnecessary waste. Nonetheless, it is 385 always a balance between sufficient power versus minimising numbers used/ sacrificed. 386 In mouse experiments, it is typical to use 5 mice for inbred strains and 8-10 for the 387 outbred, however based on our findings, a power calculation suggests that larger strains, as there is significant variation even between sub-strains based on 404 breeding history, original parent strains, and source locations (18). 405 Mouse age: JEV predominantly affects children in endemic areas, which most 406 likely reflects early life exposure leading to immunity by adulthood (16), though 407 adults are also susceptible to JE upon first exposure (17). The blood brain barrier 408 matures with age; to what extent this is relevant to human disease is not clear, but 409 in mice this is clearly relevant, with some investigators using a sham IC injection 410 in order to disrupt the blood brain barrier and allow encephalitis to develop (18). 411 The systematic review provides support for this, as mortality reduced with age. further exploration demonstrated that this was due to interaction of other variables. 437 It was notable that there was an association between the proximity of the route of 438 inoculation to the brain and mortality (p value < 0.001). 439 A refined dataset of 2562 mice was produced and a final meta-regression model run using 440 mouse age, mouse strain, virus strain, virus dose in log10PFU and route of inoculation. 441 The final model reduced the heterogeneity substantially (I^2 37.8%, df 241) such that 442 54% of the variability was explained. Despite this analysis, nearly half of the variability in 443 mouse models of JE remained unexplained, leaving significant room for variation due to 444 individual laboratories, and therefore also room for improvement in standardisation of 445 these important and useful models. of the different key variables reduces the internal validity. Nonetheless, this review still 450 represents the comprehensive body of data on mouse models of JE assembled to date. 451 We summarise our final recommendations in Table 2. Attention to performance and   452 reporting of experiments on key factors identified in this review will reduce heterogeneity 453 and enable standardisation of models. This is critical to enable evaluation of novel 454 therapeutics.  -Virus strains used in experiments need to be sequenced and the data included in publications.
-The data has been made publicly available and serves to inform future experiments in this field.