Cases of Middle East respiratory syndrome (MERS), caused by MERS-associated coronavirus (MERS-CoV), have continuously been reported since June 2012. As of 29 June 2016, the total number of laboratory-confirmed cases notified to the World Health Organization (WHO) reached 1,768 cases, including 630 deaths [1]. Particularly large outbreaks of MERS-CoV infection have been reported in the Kingdom of Saudi Arabia (KSA) and the Republic of Korea (ROK), while smaller outbreaks and importation events have been reported in other 25 countries [1]. Of these, 10 countries are located in the Middle East, 7 countries in Europe, 3 countries in Africa, 3 countries in Southeast and East Asia, and 1 in North America (the United States of America) [2–4]. Because of the regular reporting of MERS cases in the Middle East, countries across the world are now facing a continuous threat of MERS outbreak.

To understand the clinical burden of MERS, it is necessary to quantify the risk of developing severe clinical manifestations. The case fatality risk (CFR) is a measure of the risk of death among those who satisfy the case condition [5], while risks of admission to an intensive care unit (ICU) and that of requiring mechanical ventilation are also useful to measure the extent of developing severe MERS complications. However, it is not only necessary to estimate such risks, but it is also critically important to identify epidemiological determinants of those risks to then predict the risk of severe complications for each patient before the onset of disease exacerbation [6]. In previous studies, the risk of death among secondary cases was estimated based on statistical modelling and was found to range from 20 to 22%, approximately [7–11]. Meanwhile, among the primary cases, the risk of death was estimated to be greater at approximately 40%, perhaps because of biases associated with case detection and diagnosis [6–8]. As for epidemiological determinants of MERS death, elderly patients with underlying comorbidities have been identified as the most susceptible population with a high risk of death [6, 9, 11].

Despite our further understanding of the risk of developing severe MERS, the abovementioned estimates are mostly based on a subset of MERS cases; for instance, some of the risk estimates are a result of the analysis of cases diagnosed in 2015 in the ROK or KSA alone. Published articles with different study designs and populations have yielded different estimates and effect sizes associated with MERS death. Because of this variability, it is valuable to comprehensively and systematically analyze published MERS studies that have recorded the clinical prognoses of cases. A systematic review is a highly informative review method that combines published results from different studies, thereby merging and contrasting results across multiple studies and answering study questions using the pooled estimates [12]. Thus, we aimed to perform a systematic review to assess risks of death and other severe complications and determine the risk factors for MERS-associated death and contrast these results by study population and study design.

The present study was a systematic review conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [13]. PICO statement: Our study question is focused on laboratory confirmed cases of MERS regardless of their treatment status, and thus, involves only retrospective observational studies, measuring their risks of admission to Intensive Care Unit (ICU) and death and comparing those risks by age, gender and underlying comorbidities.

The present study systematically reviewed the risk of severe manifestations and death by MERS by systematically searching and analyzing published articles from the KSA and the ROK and calculating not only the CFR but also the risks of ICU admission and requiring mechanical ventilation. Several clinical predictors of death were identified including older age, male sex and underlying medical conditions, including diabetes mellitus, renal disease, respiratory disease, heart disease and hypertension. The risk estimate appeared to vary by study design. In particular, studies focusing on patients in the ICU yielded the greatest estimates, while the CFRs for surveillance and hospital outbreak studies were smaller. These findings indicate that ascertainment biases in surveillance and hospital outbreak studies, frequently involving case finding effort, were smaller than in other types of studies. The importance of case finding effort is likely reflected in the different CFR estimates based on surveillance data between KSA and ROK. Although the presently identified clinical predictors are in line with previously published studies [6, 7, 11], the present study is the first to systematically analyze published studies, including clinical research studies, and extract findings that echo those of published articles. As was observed in this study, systematic search and analysis of the transmission characteristics [38] and spatial spreading patterns of MERS [39] have been successful.

An important contribution of the present study is that we demonstrated that the risk of death or severe manifestations is highly heterogeneous for various reasons, including different study designs. It is recognized that MERS involves asymptomatic infection [9], and thus, studies must be clear as to how the risk is estimated, including the definition and diagnostic methods used to identify infected individuals. Depending on the study design, the clinical predictors of death also differed. For example, renal and heart diseases might not predict the risk of death in an ICU setting, but they may be critically important in other settings that involve milder cases. Not only studies in ICU settings, but also retrospective studies yielded relatively high risk estimates for severe manifestations and death. Our finding raises concerns regarding the retrospective analysis of confirmed cases in registered databases without referring to biases associated with case detection and diagnosis, which could yield a biased risk estimate of MERS severity. In fact, that could explain why the CFR of confirmed cases among registered cases in patients’ database has been as high as 40%, while the CFR of secondary cases in the presence of contact tracing has been estimated at about 20% [6–10].

The comorbidities identified in our study are in line with those already identified elsewhere [6, 37]. The identification of comorbidities is not only stressed based on previous and present findings [11], but it is critically important to understand the underlying pathophysiological mechanisms. High representation of men among deceased cases may reflect the interaction of factors related to sex-specific lifestyle (e.g., smoking habits in the Middle East). Older age might reflect the greater likelihood of having underlying medical conditions. Diabetes, renal and respiratory diseases could predispose patients to be immunologically vulnerable and heart disease could induce water retention (e.g., secondary aldosteronism), both exacerbating the systemic condition. Hypertension could have been confounded by some other explanatory factor (s), for example, obesity could have likely led to both hypertension and MERS death. Nevertheless, identified predictors are accompanied by reasonable biological explanations.

The present study is not free from limitations. The biggest concern is, given the absence of identifying information, the included articles most likely referred to the same cases multiple times, potentially overestimating the number of cases. In fact, the total number of diagnosed and reported cases of MERS as of June 2016 is approximately 1,768 cases, but our systematic review included as many as 2,081 cases. Thus, it is likely that multiple reports from ROK (e.g., Cowling et al. [8], KCDCP [33] and Majumder et al. [34]) reported on the same cases multiple times. Rather, we did not avoid any overlap of cases in datasets because that adjustment forced us to adjust the overlap among the cases from the KSA in a similar manner. For this reason, the pooled estimate would never represent the actual pooled outcome data because the same case was counted multiple times. If we remove Cowling et al. [8] and Majumder et al. [34] from our analysis and include KCDCP [33], which had the largest sample size, the pooled estimate of the CFR would be increased to 45.4% (95% CI: 43.2, 47.7). This is understandable owing to the diminished impact of the extensive contact tracing effort in the ROK. Despite these overlaps, we conducted this systematic review to demonstrate that ascertainment biases likely act as a key factor that characterizes differential mortality across countries. To avoid any overlap of cases and better identify risk factors of ICU admission and death, it is advised to set up a common case registration system across countries and allocate identity number for each individual case.

As the second technical limitation to remember, it should be noted that the access to individual data was not achieved, and thus, for instance the age-related analysis did not rest on individual age data, and similarly, we have had limitations in the precision of the majority of outcome evaluations. Third, clinical predictors of death have been classified only at organ level, and moreover, individual behavioral factors or habitat [40] have not been examined in relation to the risk of MERS death. Fourth, non-English language manuscripts have been missed, and they include at least a few publications in Korea and one from Jordan.

Despite these problems, we cannot help but consider that the present study successfully and systematically evaluated the risk of severe manifestations and death by MERS by collecting published information on clinical predictors of the risk of death. An important consideration is that the associated risk estimation and identification of risk factors of MERS call for particular care in terms of study design, especially in aiming to eliminate biases associated with detection and diagnosis.

Heterogeneity in risks of death and severe manifestations secondary to MERS was substantial. Differential study design was one of underlying reasons for the large heterogeneity. Statistical estimation of the risk of MERS death and identification of risk factors must be conducted with particular careful attention paid to study design, especially accounting for biases associated with case detection and diagnosis.