Revisiting MERS-CoV Outbreak Chest Radiographic Initial Findings, Temporal Progression and Correlation to Outcomes: Insights for the COVID-19 Pandemic

Amr M. Ajlan (  amrajlan@yahoo.com ) king Abdulaziz University Nesreen H. Abourokbah King Abdulaziz Medical City Samira Alturkistany King Faisal Specialist Hospital & Research Centre Ahmed Alharthy King Abdulaziz University https://orcid.org/0000-0002-2276-7988 Rayan A. Ahyad King Abdulaziz University Majed Ashour King Faisal Specialist Hospital & Research Centre Tariq A. Madani King Abdulaziz University

around the globe in symptomatic and asymptomatic individuals [6][7][8][9] . Symptomatic patients may present with respiratory symptoms, such as cough, fever and dyspnea, and non-respiratory complaints 1,6,8 . The overlap in clinical features and risk of death from either virus heightens the signi cance of such coexistence 1,8,9 . Despite the public fear of COVID-19, which has a case-fatality rate of about 1%, MERS-CoV has a higher case-fatality rate of around 34% 1,9 . Furthermore, the reported imaging magni cations of both viruses are remarkably similar [7][8][9] . The utility of chest X-rays (CXRs) in investigating MERS-CoV and COVID-19 is practically convenient, yet the number of dedicated studies addressing the CXR ndings of both coronaviruses is limited, especially when compared to that of their CT features 6,10-17 .
We retrospectively studied CXR cases from the Jeddah city initial MERS-CoV outbreak to identify the radiographic appearance on the initial CXRs and temporal progression over follow-up studies. We also studied imaging predictors of nal patients' outcomes.

Subjects:
The local research ethics committees of three participating Jeddah-located hospitals approved this retrospective study, and the consent form was waived. The electronic archives were searched from November 2013 till October 2014 for all cases with a con rmed diagnosis of MERS. Case con rmation was based on at least one positive real-time reverse-transcriptase-polymerase chain reaction (rRT-PCR) respiratory sample, by targeting MERS-CoV RNA upstream region of the E gene and the open reading frame ORF1a and ORF1b regions. Five patients with no chest radiographic imaging, and three patients younger than 18-years of age were excluded. Seven patients included in this study were previously reported, but for CT imaging features and not for CXR ndings 18 . The clinical, laboratory, and imaging ndings and outcome were analyzed, and initial false negative rRT-PCR results were recorded. Diabetes mellitus, hypertension, chronic lung diseases, cardiac diseases, chronic renal failure, neurological de cits, pregnancy or immunocompromised status were recorded as comorbidities. For outcome data analysis, patients were divided into those deceased versus those who survived; the latter include those transferred to other institutions with no further follow-up data.
Applied terminology was in line with the Fleischer Society terms 19 . Airspace opacities were classi ed as ground-glass opacities (GGO) if underlying lung markings were not obscured, as consolidation if underlying lung markings were obscured, and as mixed opacities, if both patterns coexisted. Interstitial opacities were divided into reticular if linear, nodular if rounded and reticulonodular if both patterns coexisted. The presence and laterality of pneumothoraces and pleural effusions, as well as the presence of cavities (i.e. rounded or oval air-or uid level-containing lucencies), were sought.
Lung involvement by airspace opacities was categorized as being unilateral or bilateral. The abnormality was assigned craniocaudal or transverse distributions, whenever possible, whether unilateral or bilateral, focal or diffuse. The predominant craniocaudal distribution was categorized as being in the upper lung (i.e. above the hilum), in the lower lung (i.e. below the hilum), or nonspeci c (i.e. with equal upper and lower lung involvement). The predominant transverse distribution was categorized as being central (i.e. perihilar), peripheral (i.e. subpleural), both central and peripheral (i.e. if perihilar and subpleural locations are distinctly seen on the same radiograph) or nonspeci c (i.e. scattered or diffuse opacities with no distinctive central or peripheral predominance). The abnormality was considered 'focal' if it was unilateral and predominantly con ned to the upper or lower lung; and further assigned the following locations: right upper, right lower, left upper or left lower lung.
All available initial and follow-up CXRs were evaluated for the features mentioned above. The presentation CXR (i.e. within the rst day) was designated as the 'initial' examination (period I) and was assessed separately. The follow-up radiographs were assessed according to the following periods: >1-5 follow-up days (period II), 6-10 follow-up days (period III), 11-20 follow-up days (period IV), 21-30 follow-up days (period V), 31-40, 41-50 follow-up days, (period VI) and 51-75 follow-up days (period VII). For each patient in each period, the most predominant and persistent ndings and distribution were subjectively recorded collectively for all follow-up CXRs performed during that same period. The nal follow-up chest radiograph was categorized as normal in a discharged patient, improved in a discharged patient, abnormal in an expired patient, or abnormal in a patient transferred to a non-participating isolation institution.
An overall imaging pattern was summarized for each patient in each period, from better to worse, as follows: normal (i.e. no abnormalities detected), focal opacity (i.e. unilateral airspace involvement of an upper or lower lung location), bilateral non-diffuse opacities (i.e. bilateral airspace involvement of clear upper or lower and central or peripheral predominance), or bilateral diffuse opacities (i.e. homogenous or heterogeneous bilateral airspace involvement with no clear craniocaudal or transverse predominance). Additionally, each patient was assigned a worst imaging pattern of involvement according to the period for which it was rst seen.

Statistical Analysis:
Statistical analyses were performed with the Statistical Package for Social Sciences version 21.0 for Windows (SPSS Inc., Chicago, IL, USA). Data are presented as frequencies and percentages for categorical variables, and as means ± standard deviations (SDs) for continuous variables. The correlation of initial rRT-PCR result with initial CXR patterns was analyzed using χ2 testing, with calculation of proportional agreement. Independent t-test was used to compare means and SDs of age (in years) of survived versus deceased patients. Factors of mortality were analyzed by comparing means ± SDs of age between survived and deceased patients using independent t-test, and by comparing the percentage of survived and deceased patients for other factors using χ2 or Fisher's exact tests, as appropriate. Multivariate binary regression was performed to determine whether a bilateral diffuse pattern on worst and nal CXRs -beside other covariates-was an independent factor of mortality; with results presented as odds ratios (OR) with 95% con dence intervals (CI). For all calculations, a P-value of <0.05 was considered statistically signi cant.

Patients' Results:
Patients' demographics, clinical characteristics and hospital course are detailed in Table 1, and laboratory results are detailed in Table 2. Fifty-three hospitalized patients constituted the study cohort, 33 (62%) of whom were males and 19 (36%) of whom were medical professionals ( Table 1). The age range of the total population was 23 to 76 years (mean and SD, 43.7 ± 15.4 years). Twenty-one (40%) patients had one or more comorbidity. Cough and fever were the most frequent symptoms, each encountered in 39 (74%) patients. The initial symptoms were of non-respiratory complaint in nine (17%) patients. Superadded bacterial infection occurred in 20 (38%) patients. Twenty-six (49%) patients required respiratory support during any of the studied periods. Twenty-nine (55%) patients were discharged, 15 (28%) patients died during admission, and 9 (17%) patients were transferred to non-participating institutions.
Imaging Findings: General Findings: Over the total follow-up period, a sum of 692 CXRs was performed and analyzed (range of 1-55 per patient; mean and SD, 7 ± 14 CXRs). Out of 46 patients with a known duration from symptoms onset to the time of obtaining the initial CXR, 17 (37%) had normal initial CXRs; obtained with a period of one to 34 days (mean and SD, 5.5 ± 7.7 days). According to imaged period, chest radiographic detailed ndings are summarized in Table 3, while summarized radiograph patterns, radiographic abnormalities distributions and clinical status by follow-up period are summarized in Table 4.
Evolution of CXR Imaging Findings: The subsequent imaging periods were marked by increasing frequency of bilateral involvement, reaching 30 of 39 (77%) patients imaged at period III, seven (18%) of whom had a bilateral diffuse pattern. The proportionally largest percentage of bilateral diffuse involvements was encountered during period V (seven of 22 patients; 32%) and period VI ( ve of eight patients; 63%). A bilateral non-diffuse pattern predominated otherwise in most periods, especially at periods II and V, where it constituted 46-59% of the imaged patients in each period; respectively. As for the type of follow-up ndings, mixed airspace opacities predominated throughout, reaching up to 16 (73%) of 22 patients imaged at period V. Interstitial opacities became relatively more frequent at periods III and VI, with the highest frequency being that of nine (41%) of 22 patients encountered at period V. Five (9%) of 53 patients developed a pneumothorax at different periods of the studied duration. Pleural effusions were infrequent. No cavities were detected.
Correlation of Initial rRT-PCR Results with Initial CXR Findings: The initial rRT-PCR was positive in 31 (58%) patients and negative in 21 (40%), while one (2%) patient was not tested. Out of the initial rRT-PCT samples, 17 (40%) of 43 nasopharyngeal swabs and 3 (50%) of six tracheal aspirates returned negative results. The correlation between initial rRT-PCR results (i.e. positive versus negative) and initial CXR ndings (i.e. normal versus abnormal) showed 44% proportional agreement, and there was no statistically signi cant association between initial rRT-PCR result and severity pattern on initial CXR (P-value of 0.505). Paradoxically, the percentage of abnormal initial CXRs in the initially negative rRT-PCR results group was higher than that in the positive rRT-PCR results group (14/21 or 67% versus 16/31 or 52%, respectively), but the difference was not statistically signi cant (Pvalue of 0.281). Results for this section are not presented in tables.

Predictors of Patients' Outcome:
Mortality was signi cantly associated with both worst and nal CXR patterns. That is, no mortality was observed in case of normal or focal opacity CXR patterns images in both the worst and last CXRs. However, a bilateral diffuse CXR pattern in the worst and last CXRs was associated with 62.5% and 76.9% mortality, respectively (P-value of 0.001). Additionally, patients who died were signi cantly older (>44 years) compared to survivors at follow-up (55 ±16 versus 39 ±13 patients, respectably), with a P-value of 0.001. The mortality rate was also signi cantly higher in cases with compared to cases without superadded infection (50.0% versus 15%, respectively), with a P-value of 0.011. No signi cance was elucidated for either gender or comorbidity.
On separate multivariate regression models, a bilateral diffuse CXR pattern in both worst (OR = 13, Pvalue = 0.006) and last (OR = 18, P-value = 0.002) CXRs were independently associated with patient death; alongside to an age >44 years-old, which was highly predictive of death in both groups. However, superadded infection showed no statistical signi cance on multivariant regression analysis.

Discussion
Only a few studies have addressed CXR ndings of MERS-CoV 10,15,20,21 . Our study and that of Das et al. 10 are, to the best of our knowledge, the only publications describing the temporal radiographic ndings in a relatively large hospitalized population sample. We addressed CXR imaging ndings over various periods. Compared to Das et al. 10 , we evaluated patients over a more extended follow-up period and a larger number of analyzed CXRs. We also categorized the overall CXR imaging into simple patterns, in a way that could be easier for physicians to digest and potentially correlate to outcomes.
We encountered 42% patients with initially normal CXRs, a percentage comparable to that of the study by Hamimi 15 , but different from the 17% reported otherwise 2,8,10 . In the context of the current pandemic, COVID-19 has also been reported to show initially normal CXRs in about 15-30% 8,22 .
Our cohort showed almost equal numbers of focal and bilateral non-diffuse initial CXR patterns. In line with other publications 2,8,10 , the normality of CXRs decreased as patients were followed, with an increasing number of bilateral non-diffuse or diffuse lung involvement. We noticed that the worst CXR progression occurred between a week to a month from the time of obtaining the initial CXR. GGO, whether pure or mixed with consolidation, was the most common CXR nding in our study. As the abnormalities progressed over time, superimposed consolidation appeared or progressed. Whether abnormalities were seen on initial or follow-up CXRs, the most common distribution was that of bilateral peripheral location, with or without perihilar involvement. Such ndings and distribution are another point of similarity for what is currently experienced with COVID-19 8,12,22,23 . The predilection for such peculiar distribution has previously raised the suggestion that novel viruses, from the coronavirus family or others, maybe inducing an acute organizing pneumonia reaction 18,24,25 .
Pleural effusions, pneumothoraces, cavities and bronchial wall thinking were all rarely seen in our study, and are not common in the literature of CXR imaging of MERS-CoV 2,8,10,18 . A noteworthy observation is that interstitial opacities were also uncommon in our study and other reports. However, reticulations were observed in some patients where longer-term follow-up imaging was obtained 8,18,26 . Reticulations have been noticed, although rarely, in cases of current COVID-19 infection as well 27 . There is a possibility that long-term sequelae of coronaviruses pneumonia include residual brosis 8,18 .
Avoidance of causes of false negative rRT-PCR testing has been addressed in MERS-CoV (21,25) but has not been correlated to the initial CXR appearance. In several cases in our study, rRT-PCR was initially falsely negative when the rst CXR was abnormal, and vice versa. By combining initial CXR and rRT-PCT results, the detection rate in our cohort would have increased from 59% to 87%. Of note, the issue of false negative rRT-PCR results is a current hot topic with COVID-19 pneumonia as well, for which several groups have advocated for early chest CT as a way to ameliorate challenges related to rRT-PCR diagnostic performance 23,28,29 .
Prior MERS-CoV research has shown that the risk of mortality signi cantly increases in proportion to the extent of lung involvement 2,8,10,26 . Such publications are in concordance with our observation that a bilateral diffuse pattern of lung abnormalities, as observed on the worst or nally imaged CXR, was independently related to higher death rates. On the other hand, no death occurred in patients in whose worst or nally imaged CXRs were normal or focally abnormal. Further, the risk of MERS-CoV mortality in our group was found to be highest when patients where >44 years-old. Herein lies another similarity with recent observations of COVID-19, of which older age and more substantial lung involvement are associated with poorer prognosis 8,17,30 .
The expected variability of sampled CXRs per various periods and the decreasing number of CXRs obtained as time progressed reduced the studied samples per speci c periods. Furthermore, the study size is considered relatively small for deriving generalized outcome data conclusions. This point is even more relevant when considering the potentially confounding factor of superadded infection, which may by itself, cloud CXR interpretations. Another limitation is that a proportion of patients that were eventually transferred to other institutes that were not included in the study scope were considered survivors.
To conclude, a substantial number of MERS-CoV cases may have a normal initial CXR. When abnormal, bilateral lower-lung-predominant ground-glass or mixed opacities on CXRs were the most prevalent imaging appearance. Worsening CXRs progression occurred within a week to a month from the initial imaging. Bilateral diffuse lung involvement is an independent risk for higher mortality, worsened by older patients. False negative rRT-PCR results may occur whether CXRs were initially normal or abnormal, and combining the results of both CXR and rRT-PCR provides a better diagnostic yield in patients suspected of having MERS-CoV. Finally, such insights are timely, given the various clinical and imaging similarities of MERS-CoV outbreaks with those of the COVID-19 pandemic. 4. Con ict of Interest: The authors of this manuscript declare no relationships with any companies, whose products or services may be related to the subject matter of the article.
5. Statistics and Biometry: Simple and well-known calculation were performed via SPSS. No complex statistical methods were necessary for this paper. One of the authors has signi cant statistical expertise and calculations were checked and re ned by the acknowledged Dr Mohamed Amine Haireche.
6. Informed Consent: Written informed consent was waived by the Institutional Review Boards.
7. Ethical Approval: Institutional Review Boards approval was obtained.
8. Study subjects or cohorts overlap: Seven out of the total 53 studied patient population were previously included in this study, but unlike the current paper, were evaluated for CT imaging features and not for chest radiographic ndings (such overlap has been stated on the Martials and Methods section and the prior publication PDF has been submitted).