key: cord-0232075-1rjyt87a authors: Poletto, Chiara; Pelat, Camille; Levy-Bruhl, Daniel; Yazdanpanah, Yazdan; Boelle, Pierre-Yves; Colizza, Vittoria title: Assessment of the MERS-CoV epidemic situation in the Middle East region date: 2013-11-06 journal: nan DOI: nan sha: 4016fe18801a8c01232ecdbb6767f73ef671d02c doc_id: 232075 cord_uid: 1rjyt87a The appearance of a novel coronavirus named Middle East (ME) Respiratory Syndrome Coronavirus (MERS-CoV) has raised global public health concerns regarding the current situation and its future evolution. Here we propose an integrative maximum likelihood analysis of both cluster data in the ME region and importations in Europe to assess transmission scenario and incidence of sporadic infections. Our approach is based on a spatial-transmission model integrating mobility data worldwide and allows for variations in the zoonotic/environmental transmission and underascertainment. Maximum likelihood estimates for the ME region indicate the occurrence of a subcritical epidemic (R=0.50, 95% confidence interval (CI) 0.30-0.77) associated with a 0.28 (95% CI 0.12-0.85) daily rate of sporadic introductions. Infections in the region appear to be mainly dominated by zoonotic/environmental transmissions, with possible underascertainment (95% CI of estimated to observed sporadic cases in the range 1.03-7.32). No time evolution of the situation emerges. Analyses of flight passenger data from the region indicate areas at high risk of importation. While dismissing an immediate threat for global health security, this analysis provides a baseline scenario for future reference and updates, suggests reinforced surveillance to limit underascertainment, and calls for increased alertness in high-risk areas worldwide. As of August 31, 2013, a total of 108 laboratory confirmed cases of human infection with the Health Organization (WHO) [1] . Since the emergence of the virus, a rapid coordinated response has been put in place to confront the novel epidemic emergency, through the identification and sequencing of the virus [2] , the enhancement of surveillance systems, the provision of updated information on the situation and of technical guidance for the clinical management of probable infections [3] [4] [5] [6] [7] , the identification of the possible virus reservoir [8] . There are still many uncertainties about various aspects of the outbreak, including a possible extension of the reservoir to other hosts, its full geographic extent, the path of transmission of the infection to humans and the associated risk. All these aspects call for heightened surveillance, enhanced investigations and the development and application of epidemiological methods to assess the current epidemic situation and determine its potential to spread efficiently in human population and to widely circulate at a global scale. In such situation, statistical, mathematical and computational methods allow estimating key epidemiological parameters from available data, under various assumptions and accounting for the many uncertainties. The reproductive number, i.e. the average number of secondary cases generated by a primary case, is a key summary measure of the transmissibility of an emerging infection. A first estimation of the MERS-CoV reproductive number was based on the analysis of cluster-size data with assumed cluster partition in terms of transmission trees, highlighting the similarity of current MERS-CoV situation to the prepandemic stage of the severe acute respiratory syndrome (SARS) outbreak [9] . Next to transmissibility, an additional important aspect characterizing the epidemic remains unknown, i.e. the incidence of infection. Observed cases may indeed only represent a proportion of the current epidemic, with a majority of infections going undetected because of mild illness or asymptomatic infection. This aspect also has further relevant implications for the correct estimation of other important overall statistics (e.g. the severity of the virus) and of the risk of importation of cases in other regions of the world. Limited data may also hide important changes in the virus transmissibility related, e.g., to viral adaptations to humans that may alter its pandemic potential, thus presenting an additional challenge for the situation assessment. Here we present an innovative integrative maximum likelihood approach to describe the current epidemic situation in the Middle East (ME) region and fill these gaps in current knowledge. We synthesize evidence from multiple sources of information: sizes of clusters of cases, traffic data, and imported cases outside the ME region. Methods used account for the limited information available and reporting inaccuracies. Our aim is to complete early findings on MERS-CoV epidemic by focusing on the virus transmissibility from human-to-human, its possible changes in time, the expected number of cases in the region, and the public health threat for other geographical areas based on case importation. Analytic overview. The integrative approach we use is based on a combined maximum likelihood analysis to jointly estimate the reproductive number and the daily rate of sporadic introduction of the virus in the population through zoonotic/environmental transmissions [10] . The integrative approach builds on two aspects of the currently reported outbreak -the distribution of cluster sizes, providing information on (Method 1), and the number of imported cases in countries out of the source region providing information on and based on the fit of a stochastic spatial metapopulation model integrating aviation data worldwide (Method 2 [11] . In particular we considered a Poisson offspring distribution accounting for no overdispersion around a common mean [12] and a geometric offspring distribution assuming a constant rate of transmission during an exponentially distributed infectious period [13] . As cluster sizes may be biased downwards by incomplete observation, we allowed for uncertainty by assuming that each case in a cluster would go unobserved with probability pcl during investigation ( representing no missed cases). This corresponds to the following distribution for reported cluster sizes: where is the observed size of the cluster, its real size ( ), ( | ) ∑ ( | ) , and ( | ) is the offspring distribution discussed above. Eventually, the likelihood was computed as ( We performed a sensitivity analysis by considering: (i) the complete Jordan cluster including 8 more cases (size=10) identified through a retrospective serology study carried out on 124 individuals [4, 14] ; (ii) all laboratory-confirmed cases (108) reported worldwide to WHO [1] ; (iii) laboratory-confirmed cases in the ME region up to May 31, 2013 [1] . The corresponding cluster size data are reported in Table 1 . Due to large concern around the ongoing outbreak and enhanced surveillance following the WHO guidelines for patients returning from the affected area, the detection of probable cases imported in countries out of the ME region is expected to be more complete than in the source region where primary cases may have gone undetected. Another source of information to estimate the reproductive number, discounting possible notification/surveillance biases in the source region, is therefore provided by the importation of cases in newly affected countries [15, 16] . We thus modify and extend a method already used for the estimation of the seasonal transmission potential of the 2009 H1N1 pandemic based on the calibration of a global epidemic and mobility model (GLEAM) [17, 18] to the chronology data of the pandemic invasion GLEAM is based on a spatially structured metapopulation approach comprising 3,362 subpopulations in 220 countries in the world coupled through mobility connections. The model is informed with high-resolution demographic data for 6 billion individuals and multi-scale mobility data including the full air traffic database from the International Air Transport Association (IATA) and short-range ground mobility obtained from national commuting data [18] . The infection dynamics takes place within each subpopulation and assumes a modified SEIR compartmentalization (susceptible, exposed, infectious, recovered individuals) [19] to account for different transmission scenarios in the ME region Associated confidence intervals were obtained by profiling the deviance in the 3D space [21] . It is important to note that such estimates cannot be derived from the maximum likelihood analysis of each Method considered separately, nor conditionally one to the other, and the full computation of ( |{ } { }) needs to be considered. In this respect, our integrative approach represents a substantial advance with respect to prior work based on the analysis of cluster data only [9] . Air traffic data analysis. We additionally analyzed the air traffic data integrated into GLEAM to evaluate the traffic capacity of the airports in the ME region and to assess the importation risk of the countries belonging to other regions than Western Europe. The integrated analysis led to a value equal to 0.50 (95% CI 0.30-0.77) and daily rate of MERS-CoV introductions into the human population in the ME region equal to 0.28 (0.12-0.85) ( Table 2 ). These best estimates were obtained considering a geometric offspring distribution, Table 2 ). Larger CIs but no significant variation in the parameters' estimates were observed by considering empirical data up to the end of May ( Table 2) . Analyses of traffic data expose large traffic fluxes towards the continents of Asia, Europe and Africa ( Figure 3 ) from the ME region. In addition to 6 neighboring countries of the ME region, 7 were found in South Asia that belong to the set of the first 20 countries with highest traffic from the ME region, 5 in Europe (among which the 4 countries reporting importation of cases from the affected area), and 2 in Africa. Results of our integrative modeling approach suggest the occurrence of a subcritical MERS-CoV epidemic in the ME region, as quantified by a reproductive number smaller than 1. The outbreak is not able to generate a self-sustaining epidemic in humans, and sporadic cases from zoonotic/environmental transmission are expected to represent a large fraction of the total size of the epidemic. The estimated confidence interval for the reproductive number is found to be very stable across changes in the data interpretation. In all cases, considering data up to August 31, we found that it is highly unlikely (<5% probability) to have a MERS-CoV outbreak with above 0.80 or below 0.30. The variation of the best estimate from the baseline case ( =0.50) to the various scenarios explored as sensitivity analysis (up to =0.69) is explained by the presence of a large region in the parameter space ( ) where the likelihood function shows small variation around its maximum value (darker red area in Figure 2 ). This is likely induced by the limited data available not allowing us to narrow down the confidence intervals of the estimates. The analysis based on the integration of two independent methods allows us to provide an estimate for the daily rate of introductions of MERS-CoV infections in human population in the ME region, in addition to the estimate for the reproductive number. The estimated 95% CI in the baseline scenario (0.12-0.85) compared to the observed value (0.116) suggests a negligible to significant underascertainment rate for zoonotic/environmental transmissions (1.03-7.32 times the reported sporadic cases), indicating that notified sporadic cases likely represent a substantial proportion of the total, but improved surveillance in the region including serological surveys around cases is needed. Since evidence for mild illness, as well as for a wide spectrum of clinical disease, was observed [3, 6] , our findings are compatible with an underascertainment rate for zoonotic/environmental transmissions that may be due to a selection bias towards more severe cases, where patients having mild illnesses or asymptomatic infections may go undetected [22] . Our estimates for the reproductive number are consistent with the results of Ref. [9] -the only study to date reporting results on interhuman transmissibility -thus further confirming the robustness of our epidemic assessment. Our work presents however substantial differences in the methodology and in its achievable predictions that we discuss in the following. One major difference is that our integrative approach also allows the quantification of sporadic cases underascertainment through the estimate of the rate of introduction of sporadic cases in the ME region, combined with the estimate for the reproductive number. In [9] , daily introductions are simply calculated on the basis of the two assumed scenarios for the transmission trees, i.e. from the assumed number of index cases among the reported data. Our procedure instead makes no assumption on the completeness of reported data, or on the local transmission trees, and relies on alternative data sources (case importations) to estimate the number of sporadic cases in the region. The cluster data analysis of Method 1 relies on the assumption that each cluster is the result of human-to-human transmission starting from a single index case. While we allow for uncertainty in case detection in the close contact investigation, we do not consider the possibility of coexposure of epidemiologically linked cases to the same source of zoonotic/environmental infection, differently from [9] . Other events that are related to human movements and mixing may as well alter the assessed scenario. Vast international mass gatherings to be taking place in the Kingdom of Saudi Arabia in the upcoming months are expected to bring large number of pilgrims to the affected area, with increased rates of local mixing that may favor the transmission of the virus, followed by a potential amplification of its international dissemination due to the return of pilgrims to their own countries [24] . It should be noted however that the Hajj 2012 occurred with an ongoing MERS-CoV outbreak in the ME region, and that MERS-CoV was absent among French pilgrims screened prior to returning to France after their participation to the pilgrimage [25] . Similarly, no increase in case notification occurred following the Umra pilgrimage in July 2013. Additional studies in travelers next to enhanced local surveillance in the region and guidance to local authorities would help to assess and control possible changes in the virus transmission with respect to last year's experience [26] . Air travel clearly represents the main mean for global spatial spread of infectious disease epidemics in the modern world, as it was previously experienced with SARS and the 2009 H1N1 pandemic [15, 16, [27] [28] [29] . Besides seasonal variations due to specific events (e.g. mass gatherings) or in/out flows of expats for seasonal jobs, a potential emerging pandemic in the ME area would constitute a very high risk for considerably rapid and wide international spread. The Figure 3 ). Similar results were also reported in Ref. [24] . Should the outbreak evolve in a self-sustained epidemic, such risk assessment analyses cannot rely on travel data only and would require the full integration of the air travel data with an epidemic model, as in GLEAM, to explicitly simulate the evolving epidemic, estimate importation likelihood [28] and provide predictions for future stages of the epidemic [16, 31] . With a subcritical epidemic in the ME region associated with a large potential for international dissemination, priority for the epidemic control should be given to the identification of the transmission of infection to humans to limit sporadic cases, to the reduction of human-to-human transmission through rapid case identification and isolation, and to the enhancement of surveillance systems in those countries that are at a higher risk of importation because of travel flows to/from the affected area. a to be comparable with the other estimates, this value has been rescaled to take into account the change of population size of the seed region; it thus represents the daily rate of sporadic cases scaled to the ME region. Figure 1 . Schematic representation of the integrative maximum likelihood approach. Method 1 (bottom circle) is based on the maximum likelihood analysis of cluster size distribution obtained from laboratory-confirmed cases in the ME region (countries in red in the zoomed area). Method 2 (top panel) is based on the maximum likelihood analysis on data on case importations in countries in Western Europe, as schematically indicated on the map. For each point in the parameter space ( ) we run 4,000 stochastic GLEAM simulations from the same initial conditions and parameterized as described in the main text. With each run providing the simulated number of imported cases for a given country , we can compare the resulting simulated probability distribution of with the observed value for that country and compute a likelihood function for all countries in Western Europe. Global Alert and Response (GAR), Coronavirus infections Detection of a novel human coronavirus by real-time reverse-transcription polymerase chain reaction Health Protection Agency (HPA) UK Novel Coronavirus Investigation team Evidence of person-to-person transmission within a family cluster of novel coronavirus infections epidemiological findings from a retrospective investigation Clinical features and viral diagnosis of two cases of infection with Middle East respiratory syndrome coronavirus: a report of nosocomial transmission Hospital outbreak of Middle East respiratory syndrome coronavirus Contact investigation of a case of human novel coronavirus infection treated in a German hospital Middle East respiratory syndrome coronavirus in bats, Saudi Arabia. Emerg Infect Dis Interhuman transmissibility of Middle East respiratory syndrome coronavirus: estimation of pandemic risk European Centre for Disease Prevention and Control. Rapid Risk Assessment: Severe respiratory disease associated with Middle East respiratory syndrome coronavirus (MERS-CoV) Branching process models for surveillance of infectious diseases controlled by mass vaccination Farrington CP Epidemiology of transmissible diseases after elimination Public health risk from the avian H5N1 influenza epidemic Jordan, retrospective case identification, WHO, request for information Pandemic potential of a strain of influenza A(H1N1): early findings Seasonal transmission potential and activity peaks of the new influenza A(H1N1): a Monte Carlo likelihood analysis based on human mobility Modeling the Worldwide Spread of Pandemic Influenza: Baseline Case and Containment Interventions Multiscale mobility networks and the large scale spreading of infectious diseases Infectious Diseases of Humans: Dynamics and Control A Novel Coronavirus Called "MERS-CoV" in the Arabian Peninsula A method for computing profile-likelihood based confidence intervals MERS-CoV update as of 26 Adaptation of SARS coronavirus to humans Potential for the International Spread of Middle East Respiratory Syndrome in Association with Mass Gatherings in Saudi Arabia Lack of nasal carriage of novel corona virus (HCoV-EMC) in French Hajj pilgrims returning from the Hajj 2012, despite a high rate of respiratory symptoms World Health Organization, International travel and Health, World-travel advice on MERS-CoV for pilgrimages Forecast and control of epidemics in a globalized world Predictability and epidemic pathways in global outbreaks of infectious diseases: the SARS case study Spread of a Novel Influenza A (H1N1) Virus via Global Airline Transportation International Civil Aviation Organization and Civil Aviation Statistics of the World Real-time numerical forecast of global epidemic spreading: case study of 2009 A/H1N1pdm The authors would like to thank J-C Desenclos and S Deuffic-Burban for useful interactions and comments. This work is partly supported by the ERC Ideas contract no. ERC-2007-Stg204863 (EpiFor) to ChP and VC; the EC-Health contract no. 278433 (PREDEMICS) to ChP and VC; the ANR contract no. ANR-12-MONU-0018 (HARMSFLU) to DLB and VC. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.