Marburg Virus Disease outbreaks, mathematical models, and disease parameters: a Systematic Review

Recent Marburg virus disease (MVD) outbreaks in Equatorial Guinea and Tanzania highlighted the importance of better understanding this highly lethal infectious pathogen. We conducted a systematic review (PROSPERO CRD42023393345), reported according to PRISMA guidelines, of peer-reviewed papers reporting historical outbreaks, modelling studies and epidemiological parameters focused on MVD. We searched PubMed and Web of Science until 31/03/2023. Two reviewers evaluated all titles and abstracts, with consensus-based decision-making. To ensure agreement, 31% (13/42) of studies were double-extracted and a custom-designed quality assessment questionnaire was used for risk of bias assessment. We present detailed information on 478 reported cases and 385 deaths from MVD. Analysis of historical outbreaks and seroprevalence estimates suggests the possibility of undetected MVD outbreaks, asymptomatic transmission and/or cross-reactivity with other pathogens. Only one study presented a mathematical model of MVD transmission. We estimate an unadjusted, pooled total random effect case fatality ratio for MVD of 61.9% (95% CI: 38.8-80.6%, I2=93%). We identify important epidemiological parameters relating to transmission and natural history for which there are few estimates. This review and the accompanying database provide a comprehensive overview of MVD epidemiology, and identify key knowledge gaps, contributing crucial information for mathematical models to support future MVD epidemic responses.


A.1 Study selection
Original research papers in English were included if reporting on MVD transmission, evolution, natural history, severity, seroprevalence, size of previous outbreaks or published mathematical transmission models.Non-peer reviewed literature was excluded.Papers identified in the search were imported into Covidence, a software program used to manage systematic reviews.From a team of seven reviewers, two independent reviewers first screened titles and abstracts then full texts to assess eligibility for data extraction.Disagreements in eligibility determination were resolved by consensus between the two independent reviewers.

A.2 Data Extraction
Outbreaks Possible methods of case confirmation included rapid diagnostic tests (RDTs) or polymerase chain reaction (PCR) tests.

Models
Details of models of disease transmission were extracted, including model type, whether it modelled deterministic or stochastic processes, whether the model was theoretical or fitted to data, availability of model code, and in the case of a compartmental model, model subclassification (e.g., SIR, SEIR).Finally, we extracted information on transmission routes modelled, underlying model assumptions, and interventions included within the model.

Parameters
Survey context of parameter values included survey location and dates, sample size, basic demographic information, and timing of the survey in relation to reported outbreaks.For each parameter, we extracted all available information, including types of values (e.g., mean, standard deviation, median), uncertainty intervals (capturing the precision of estimates), and ranges (if when multiple estimates were obtained from different populations or using different methods).
We recorded the methods used for reproduction number estimation (e.g.renewal equations, empirical methods, compartmental models).For case fatality ratios, we extracted whether the estimation approach accounted for cases with unknown final status or not.For case fatality ratios and seroprevalence, we additionally recorded numerators and denominators where available.For genomic data, we noted the gene studied if specified and if the sequence data were available.
Outcome (e.g.infection or death), the risk factor for that outcome (e.g.age, sex or occupation), the type of occupation if specified, and whether the risk factor(s) estimates were statistically significant and/or adjusted.We chose not to extract odds ratio estimates because studies may have used different stratifications or reference groups, making it challenging to compare values across studies.The information we extract offers an overview of risk factors explored across studies that may affect the risk of infection and death, which may be useful to consider when designing MVD transmission models.

A.3 Analysis
We presented the de-duplicated start and end dates, deaths, confirmed, suspected, asymptomatic, and severe/hospitalised cases, and method of case confirmation in Table B , where n denotes the number of cases (e.g. the denominator used to derive the CFR estimate).
The mean generation time of 9.15 days is computed as the simple average of the two extracted generation time parameters from Ajelli et al [1] of 9 days and 9.3 days.

A.4 Meta-analysis methods
We provide a brief overview of the methodology used in the meta-analysis of CFRs in Figure ??.We followed a standard methodology for systematic reviews, and all analyses were performed using the meta R package [2].A mixed effects model is a linear model such that y i = β 0 + j β j x i + u i + ϵ i , where β = (β 0 , ..., β j ) ⊤ represent the fixed effects, y i the observed data, x i any explanatory variable and u i are the random effect terms, centred around zero and independent across i, and ϵ i are error terms.Meta-analysis is a special case of the above mixed-effects model with only an intercept term (β 0 ) and a random-effects term u i associated with that intercept.
We derived a CFR for each study, either using the CFR directly as reported in the paper or by calculating the unadjusted CFR from outbreak data (deaths/cases).We then transformed the CFRs using logit-transformation y i = log CF Ri 1−CF Ri to ensure that the distribution was approximately normal.Finally, we used a generalised logistic mixed-effects model with the transformed CFRs as the outcome to estimate the pooled effect.
A comprehensive overview of the methodology is provided by [3].

A.5 Quality Assessment
Possible responses for each question listed below were Yes, No, and Not applicable.

Theme Question
Is the methodological/statistical approach suitable?(how the data are used)

Non-English language publication
Mention of historical or any outbreak in humans: size, year, location, duration, spatial scale Studies of co-infections.(local, regional, national, international).
Measures/estimates of animal: R, R 0 , Rt, r, Re, growth rate, mutation rate.
Animal studies that do not report R, Rt etc.

Mathematical or statistical model of transmission.
Qualitative studies, e.g., KAP studies.
Measures of seroprevalence and negative seroprevalence in humans.
Pathogen not the primary focus of study.
Relative ratio of human-human vs animal introductions. Duplicates.
Reviews that report inclusion criteria for reference checking.Does not match any of the inclusion criteria.
For "small" pathogens, include case reports to potentially reconstruct serial interval distribution etc.
In-vitro studies.
Non-peer reviewed publications, conference proceedings, abstracts, posters, letters to the editor Papers that reference the city of Marburg in Germany instead of MVD     Risk factors were mapped onto our risk factor classification (see Supplement) by interpreting the authors' descriptions.Adjusted refers to whether estimates were adjusted (i.e. from a multivariate analysis) or not (i.e. from a univariate analysis), with unknown showing that this information is not clearly stated in the original study.Statistical significance was determined according to the original authors' statistical approaches when specified, or using a p-value of 0.05 otherwise.The numbers in the significant and not significant columns represent the total sample size included in the analyses for this risk factor and outcome category.

C epireview
We developed an R package called epireview that provides a central location to host and access the extracted data for the nine priority pathogens, allows for submissions of outbreak, model, and parameter data from new peer-reviewed papers via pull requests, and includes functions to produce the figures and tables included in this paper and update them with any additional data.This package will be updated as the overall project by the Pathogen Epidemiology Review Group (PERG) continues to extract modelling parameters for the rest of the nine priority pathogens as defined by WHO.
There are several vignettes available: • A vignette for MVD with tables and figures from this paper that will be updated as data are added to the database.
• A vignette that lists the options for each model, outbreak, or parameter field and describes how to access them using a function in the package.
• A vignette to explain the process of updating the database with new article, model, outbreak, or pathogen data.

Results
Study selection 16a Describe the results of the search and selection process, from the number of records identified in the search to the number of studies included in the review, ideally using a flow diagram.page 3 16b Cite studies that might appear to meet the inclusion criteria, but which were excluded, and explain why they were excluded.

Figure B. 1 :#
Figure B.1: The number of each type of parameter extracted from studies included in the review.

Figure B. 2 :
Figure B.2: Overview of the estimates of the case fatality ratio (CFR) obtained from the included studies.(A) CFR estimates reported in the included studies, stratified according to estimation method.Points represent central estimates.Error bars represent an uncertainty interval associated with the point estimate, as reported in the original study.(B) CFR estimated from extracted outbreak data, including only one observation per outbreak using the study with the longest duration of the outbreak reported ensuring each case is not double counted.Shaded bars represents the imputed binomial confidence interval for studies with a sample size, n > 1. Vertical dotted lines represent 0% and 100% CFR.

Figure B. 3 :
Figure B.3: (A) Count of papers for each quality assessment question scoring Yes, No or not applicable.(B) Quality Assessment Score defined as proportion of Yes votes for each paper relative to sum of Yes and No answers, removing NAs.The time trend is fitted using Local Polynomial Regression Fitting.

Figure B. 4 :
Figure B.4: More detailed table of risk factor data from the extracted studies, giving countries, times and contexts of surveys and non-aggregated information on each risk factor assessed in the four relevant studies.
.3.Confidence intervals for the unadjusted CFR estimates, as shown in Figure B.2, were computed as ( ĈF R ± 1.96 ×

Table B
.2: Inclusion and exclusion criteria.

Table B
.3: Outbreak form fields. Refer to epireview in Supplement C for dropdown options.

Table B .
4: Model form fields. Refer to epireview in Supplement C for dropdown options.Dates of survey in format YYYY, YYYY-YYYY, MMM YYYY, or MMM-MMM YYYY from survey start and end variables Table B.5: Parameter form fields. Refer to epireview in Supplement C for dropdown options.

Table B
.6: Overview of the MVD delay parameter estimates extracted from the included studies.These are stratified into five categories: Generation Time, Incubation Period, Time in Care, Time from Symptom to Careseeking and Time from Symptom to Outcome.Estimates and associated uncertainty are provided, along with information regarding the population group corresponding to the estimate and the timing and location of the outbreak.'Other'refersto a range of different values which are specified in the underlying papers.Risk factorAdjusted Sample size (Significant) Sample size (Not significant)

Table B
.7: Aggregated information on risk factors associated with MVD infection and seropositivity.
Present results for main outcomes, preferably indicating the number of included studies and participants for each.If meta-analysis was done, report the summary estimate and confidence/credible interval.If comparing groups, indicate the direction of the effect (i.e. which group is favoured).Specify the methods used to decide whether a study met the inclusion criteria of the review, including how many reviewers screened each record and each report retrieved, whether they worked independently, and if applicable, details of automation tools used in the process.
9Provide a brief summary of the limitations of the evidence included in the review (e.g.study risk of bias, inconsistency and imprecision).6Specifyalldatabases,registers,websites,organisations, reference lists and other sources searched or consulted to identify studies.Specify the date when each source was last searched or consulted.page2Searchstrategy7Presentthefull search strategies for all databases, registers and websites, including any filters and limits used.page 2 + Figure 1 Table D.9: PRISMA 2020 Checklist.([4])13bDescribeany methods required to prepare the data for presentation or synthesis, such as handling of missing summary statistics, or data conversions.page2/313cDescribe any methods used to tabulate or visually display results of individual studies and syntheses.page2/313dDescribeany methods used to synthesize results and provide a rationale for the choice(s).If meta-analysis was performed, describe the model(s), method(s) to identify the presence and extent of statistical heterogeneity, and software package(s) used.page313e Describe any methods used to explore possible causes of heterogeneity among study results (e.g.subgroup analysis, meta-regression).

Table F
Systematic reviews were excluded from our analysis, but TableE.10 presents the systematic reviews which we have found and used for validation purposes.We list all excluded studies with reasons for exclusion in TableF.11 Table D.9: PRISMA 2020 Checklist.([4]) 27 Report which of the following are publicly available and where they can be found: template data collection forms; data extracted from included studies; data used for all analyses; analytic code; any other materials used in the review.page 9 Table D.9: PRISMA 2020 Checklist.([4]) E Systematic Reviews Table F.11: Excluded studies at full text review with exclusion reason

Table F
.11: Excluded studies at full text review with exclusion reason

Table F .
11: Excluded studies at full text review with exclusion reason Table F.11: Excluded studies at full text review with exclusion reason

Table F .
11: Excluded studies at full text review with exclusion reasonTable F.11: Excluded studies at full text review with exclusion reason Table F.11: Excluded studies at full text review with exclusion reason Not peer-reviewed paper; JS (2019-08-08 23:55:45)(Select): This seems to not really be a journal article, but a bunch of summaries?Probably not a peer-reviewed jour-Table F.11: Excluded studies at full text review with exclusion reason Table F.11: Excluded studies at full text review with exclusion reason

Table F .
11: Excluded studies at full text review with exclusion reason