Rapid implementation of SARS-CoV-2 emergency use authorization RT-PCR testing and experience at an academic medical institution

coronavirus disease 19 (COVID- 19), New York City, real-time reverse 14 transcription-polymerase chain reaction (rRT-PCR), severe acute respiratory syndrome 15 coronavirus 2 (SARS-CoV-2) 16 17 18 *Correspondence to: Hanna Rennert 19 Department of Pathology and Laboratory Medicine 20 Weill Cornell Medicine 21 545 East 68 Street, New York, NY 10065 22 Email: [email protected] 23 24 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice. 2

25 An epidemic caused by an outbreak of s evere acute respiratory syndrome coronavirus 2 26 (SARS-CoV-2) in China in December 2019 has since rapidly spread internationally, requiring 27 urgent response from the clinical diagnostics community. We present a detailed overview of 28 the clinical validation and implementation of the first laboratory-developed real-time reverse-29 transcription-PCR (rRT-PCR) test offered in the NewYork-Presbyterian Hospital system 30 following the emergency use authority ( EUA) guidance issued by the US Food and Drug 31 Administration. Validation was performed on nasopharyngeal and sputum specimens 32 (n=124) using newly designed dual-target rRT-PCR (altona RealStar® SARS-CoV- 2 33 Reagent) for detecting of SARS-CoV-2 in upper respiratory and lower respiratory tract 34 specimens, including bronchoalveolar lavage and tracheal aspirates. Accuracy testing 35 demonstrated excellent assay agreement between expected and observed values. The limit 36 of detection (LOD) was 2.7 and 2 3.0 gene copies/reaction for nasopharyngeal and sputum 37 specimens, respectively. Retrospective analysis of 1,694 tests from 1,571 patients revealed 38 increased positivity in older patients and males compared to females, and an increasing 39 positivity rate from approximately 20% at the start of testing to 50% at the end of testing three 40 weeks later. Our findings demonstrate that the assay accurately and sensitively identifies 41 SARS-CoV-2 in multiple specimen types in the clinical setting and summarizes clinical data 42 from early in the epidemic in New York City. 43 44 45 46 47 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 3

48 The novel coronavirus SARS CoV-2 is a member of the Betacoronavirus genera in the 49 subfamily Coronavirinae, which is known to cause respiratory illness and gastroenteritis 50 in humans and other mammals [1, 2]. Two other Betacoronavirus that have met with 51 global attention are SARS-CoV (2002) and MERS-CoV (2012). An outbreak of 52 respiratory disease caused by SARS-CoV-2, first detected in Wuhan, China at the end of 53 December 2019, rapidly spread to other countries, including the United States [3, 4], 54 resulting in New York City in particular becoming an epicenter of the pandemic [5]. Given 55 the devastating impact on the healthcare system and the need for accurate and quick 56 diagnosis of SARS-CoV -2 infection the United States Food and Drug Administration 57 (FDA) has established a rapid pathway for using laboratory-developed tests (LDTs) that 58 was outlined in a guidance document published on February 29, 2020 [6] . According to 59 this guidance SARS-CoV-2 testing may be performed by CLIA-certified high-complexity 60 molecular laboratories under Emergency Use Authorization (EUA) according to a set of 61 recommendations regarding the minimum validation required for ensuring the analytical 62 and clinical validity of the test. Details of the test and validation must be submitted by the 63 laboratory to the FDA through an EUA application within 15 days of initiating testing, after 64 which testing may continue provisionally until a decision by the FDA is rendered. 65 66 The Center for Disease Control (CDC) and the New York State Department of Health 67 (NYS DOH) had designed and manufactured new test kits for SARS -CoV-2. However, 68 very few laboratories were able to get access to these reagents , and their use has been 69 limited by the need for specific instruments, which were not available in our institutio n. 70 Additionally, limited access to SARS-CoV-2 RNA reference control material presented a 71 significant hurdle to the validation process. The FDA EUA announcement allowed 72 laboratories to procure SARS-CoV-2 RNA from the World Reference Center for Emerging 73 Viruses and Arboviruses (WRCEVA) or the National Institutes of Health (NIH) Biodefense 74 and Emerging Infections Research Resources Repository (BEI). 75 76 The scale of demand for diagnostic testing and the shortage of supplies led to the need 77 for a high throughput diagnostic test that could be readily implemented in a variety of 78 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 4 laboratories. Here we describe how in less than five days, our institution validated and 79 submitted for FDA EUA approval the research use only (RUO) RealStar® SARS-CoV-2 80 Reagent Kit 1.0 (Altona Diagnostics) test. Comparable validation studies that can take 81 months were completed with in a week. We also detail workflow considerations and 82

from three weeks of testing from March 11, 2020, through March 31, 2020, during 83 the early days of the COVID-19 outbreak in New York City. 84 85

86 SARS-CoV-2 RNA control material 87 We obtained SARS-CoV-2 RNA reference material from WRCEVA (University of Texas 88 Medical Branch, Galveston, TX, Strain USA_WA1/2020, Lot TVP 23156, RNA 89 preparation date 2/21/2020) for use in clinical evaluation and limit of detection (LOD) 90 studies. We used this RNA reference material to perform LOD dilution series experiments 91 and to create contrived positive samples for accuracy studies by spiking it into pooled 92 leftover negative patient specimens. 93 94 Validation samples and clinical cohort 95 An in-house validation panel consisting of a total of 124 contrived s amples and patient 96 specimens, including NP ( 64) and sputum (60) specimen types, was used for the 97 validation. Samples were obtained from individuals suspected of respiratory tract 98 infections. All N P samples had been clinically tested for the presence of twenty -one 99 common respiratory viruses using our institution’s respiratory virus panel, the 100 commercially available BioFire FilmArray® Respiratory Pathogen 2 (RP2) panel (BioFire 101 Diagnostics, LLC, Salt Lake City, USA). Reactive c linical samples consisted of four 102 patient specimens confirmed to have SARS-CoV-2 by the New York City-Department of 103 Health and Mental Hygiene (NYC- DOH) using the NYS-DOH SARS-CoV-2 EUA assay 104 and samples contrived by spiking WRCEVA RNA material into pooled leftover negative 105 clinical specimens. 106 107 Additionally, we performed a retrospective analysis of patient characteristics on 1,694 108 consecutive upper respiratory tract (URT) specimens tested on the assay obtained from 109 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 5 1,571 patients with high suspicion for COVID- 19 that were treated at NewYork 110 Presbyterian Hospital (NYPH) campuses from March 11 to March 31, 2020 . The IRB 111 Committee at Weill Cornel Medicine (WCM) approved this study. 112 113 Real time reverse-transcription PCR testing 114 Automated extraction of total nucleic acid (TNA) was performed on 200 L of NP swab 115 viral transport media (VTM) following an off-board lysis viral inactivation step, using the 116 QIAsymphony DSP Virus/Pathogen Mini Kit coupled on the QIAsymphony SP (Qiagen, 117 Germantown, MD), with a resulting eluate volume of 60 L. The viral inactivation step 118 was performed in a class 2 biosafety cabinet using personal protective equipment 119 following our hospital biosafety policies. For sputum, 100 L of specimen was first 120 treated with 0.3% dithiothreitol (DTT) solution (1:1 ratio) and incubated at 37° C for 30 121 minutes to reduce viscosity. One-step reverse transcription to cDNA and rRT- PCR of 122 viral targets (Envalope and Spike genes) and internal control (IC) was performed using 123 10 L TNA eluate and the RealStar ® SARS-CoV-2 real-time RT-PCR Kit 1.0 (altona 124 Diagnostics Gmbh, Hamburg, Germany) on the Rotor-Gene Q Thermocyler (Qiagen) for 125 a total volume of 30 L per reaction. PCR amplification and detection were performed 126 using multi-color fluorescent dye-labeled probes for the identification and differentiation 127 of B-betacoronavirus (B-βCoV) and SARS-CoV-2 specific RNA and the detection of the 128 IC within one reaction, allowing for higher throughput testing compared to the CDC assay. 129 Samples in which both the E gene target (all B-βCoV) and the S gene target (SARS-CoV-130 2 specific), or the S gene target only were detected within the first 40 cycles of 131 amplification were considered “Detected”. Samples with cycle threshold (Ct) values > 132 40.0 were considered negative. Each run contained an external positive control (1:10 3 133 WRCEVA RNA dilution), positive kit control (synthetic B-ßCoV and SARS-CoV-2 RNA), 134 SARS-CoV-2 NP negative control, and a non-template control (NTC). 135 136 Assay Performance characteristics 137 The FDA EUA mechanism specified four distinct performance characteristics consisting 138 of limit of detection (LOD), inclusivity (analytical sensitivity) cross -reactivity (analytical 139 specificity), and clinical evaluation (accuracy) studies. For the LOD studies, SARS-CoV-140 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 6 2 inactivated virus or RNA spiked into artificial or real clinical matrix was acceptable, as 141 long as the matrix was from the most difficult specimen type accepted for testing on the 142 clinical assay (in decreasing order of difficulty : sputum, other lower respiratory tract 143 specimens, and N P or oropharyngeal (OP) swabs collected and transported in viral 144 transport media). NP and sputum samples were tested on the altona RealStar® rRT-PCR 145 assay to ensure the absence of SARS-CoV-2 and pooled for use as a matrix for spiking 146 in RNA for LOD studies and accuracy studies. Six ten-fold serial dilutions (1 x 101 to 1 x 147 107) were performed with three replicates at each concentration by spiking WRCEVA 148 RNA reference material (60,000 pfu/L stock WRCEVA RNA reference material; ~6:107 149 genomic copies/L) into NP and sputum eluates obtained from pooled -negative patient 150 NP or sputum specimens. The dilutions at the LOD were performed by spiking 1 L of 151 1:105 dilution to 60 L eluate , yielding a concentration of 0.01 pfu/uL=10pfu/mL 152 (~10,000copies/mL). Assay performance at the determined LOD was confirmed with at 153 least 20 additional replicates for each type of sample (sputum and NP). 154 155 For the accuracy studies, a total of 6 4 positive (34 NP and 30 sputum specimens, 156 respectively) specimens and 60 negative (30 NP swabs and 30 sputum) specimens that 157 tested negative for SARS -CoV-2 were used. Positive specimens were either contrived 158 positive samples generated by spiking WRCEVA RNA reference material into pooled 159 leftover negative NP or sputum specimens or real patient specimens, as described above. 160 Twenty of the contrived clinical specimens were spiked at a concentration of 1x-2x LoD, 161 with the remainder of s amples spanning the assay testing range. FDA defines the 162 acceptance criteria for the performa nce as 95% agreement at 1x -2x LOD, and 100% 163 agreement at all other concentrations and negative specimens [6]. 164 165 Inclusivity and cross -reactivity studies used a combination of in silico and in vitro 166 approaches. As the primer and probe sequences were proprietary to the kit 167 manufacturer, we included the results of their in silico analysis in our EUA application. 168 Additional studies to determine cross -reactivity were performed in our laboratory by 169 testing 10 NP samples that were positive by the RP2 for the four human coronaviruses 170 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 7 NL63 (n=2), 229E (n=2), OC43 (n=4), or HKU1 (n=2), defined as high priority pathogens 171 from the same genetic family by the FDA. 172 173 Data analysis and statistical methods 174 Data analyses, including statistics and plot generation, were performed using R 175 programming language v 3.6.0 [7]. LOD was determined through a probit regression 176 model using the glm function following CLSI EP17A2E Guidance with Application to 177 Quantitative Molecular Measurement Procedures [8]. 178 179

180 Validation of Assay 181 a. Limit of detection 182 Dilution series studies on pooled negative N P specimens spiked with WRCEVA RNA 183

material, with three replicates across a viral range of 1 gene copy/reaction to 184 1,000,000 gene copies/reaction (1 to 6 log10), demonstrated an accurate and linear 185 response across five logs of detection for NP and four logs of detection for sputum (Table 186 1, Figure 1). Probit analysis was applied to the NP data after an additional five replicates 187 of testing were performed at 0.8, 0.6, 0.5, 0.4, and 0.2 gene copies/reactio n, and 188 narrowed the LOD to 2.7 gene copies/reaction at 95% detection rate (Figure 2). A similar 189 LOD series and probit analysis was performed on sputum at 80, 60, 50, 40, and 20 gene 190 copies/reaction, and resulted in a lower sensitivity with a LOD of 23.0 gene 191 copies/reaction at 95% detection rate in sputum compared to N P specimens. For both 192 NP and sputum specimens, 20/20 and 23/23 additional replicates tested at the ir LODs, 193 respectively, resulted as positive. 194 195 b. Inclusivity and Specificity 196 The in silico analysis for inclusivity that was performed by the manufacturer of the kit 197 found 100% homology of the E gene and S gene forward and reverse primers and probes 198 with 563 whole-genome sequences of SARS-CoV-2 published in GISAID and NCBI as of 199 3/16/2020 [9]. The in silico analysis for cross -reactivity that was performed by the 200 manufacturer found that the S gene and E gene forward and reverse primers and probes 201 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 8 had less than 80% homology with the vast majority of 40 different pathogens (125 strains 202 total) tested (see Supplementary Material). In cases with greater than 80% homology, 203 cross-reactivity was not a concern as only the forward or reverse primer, but never both 204 primers were affected , thus rendering amplification impossible. Additionally, all ten 205 samples that had human coronaviruses NL63, 229E, OC43, or HKU1 detected by RP2 206 panel tested negative for SARS-CoV-2 on the RealStar® rRT-PCR assay. 207 208 c. Clinical evaluation-Accuracy 209 All 30 NP specimens that tested negative on the RP2 panel also tested negative on the 210 SARS-CoV-2 rRT-PCR assay (Table 2). Leftover VTM from these negative specimens 211 was pooled to create a sample matrix for the LOD and contrived positive sample studies. 212 Clinical evaluation studies resulted in the detection of SARS-CoV-2 in all specimens 213 contrived by spiking WRCEVA RNA reference material (n=20) into pooled SARS-CoV-2 214 negative NP VTM or sputum and all four positive patient samples tested by NYC -DOH 215 (Table 2 and Supplementary Table 1). The high -positive patient sample run at 216 successive dilutions (n=10) remained positive throughout the range of concentrations (1:2 217 to 1:1,024; Ct range, 22-31) (Figure 1). Similar clinical evaluation studies were performed 218 for sputum specimens ( Table 2 and Supplementary Table 2 ), also with 100% 219 concordance. Validation studies performed on bronchoalveolar lavage ( BAL) and 220 tracheal aspirate specimens and additional sample collection systems (Table 3 and 221 Supplementary Material including Tables 3 and 4) showed 100% accuracy. 222 223 Clinical cohort characterization 224 The Altona rRT-PCR SARS-CoV-2 test was used to test N P and OP swabs from March 225 11, 2020, to March 30, 2020. Starting March 30, 2020, our institution deployed the higher 226 throughput Roche cobas 6800 SARS -CoV-2 rRT-PCR test (Roche Molecular Systems, 227 Inc; Branchburg, NJ) [10] to meet increasing specimen volumes. During the initial phase 228 with only Altona testing, 1,694 tests were performed on 1,372 NP (40% positive), 57 OP 229 (19% positive), and 311 NP/OP (25% positive) swab specimens from 1,571 patients. Ct 230 values were not significantly different for the E gene, S gene, and I C targets between 231 positive NP, OP, or NP/OP samples ( Supplementary Figure 1 ). The number of tests 232 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 9 with indeterminate or B-βCoV results were 5 and 4, respectively, all in NP swab samples. 233 The mean Ct for E gene, S gene, and I C targets in positive samples were 23.0 (11.1-234 40.7), 22.5 (10.3-40.6), and 29.6 (27.0-38.4), respectively (Figure 3). Using Ct value of 235 the S target gene as a surrogate for viral burden, the upper respiratory tract specimens 236 could be classified into three groups: high (Ct 30, n = 89, 13.8%). Over three weeks of testing, more 238 than 75% of positive samples could be classified as having medium to high viral burden. 239 240 Of 135 patients with repeat testing , only 17 had different results on the second test 241 including 13 patients who first tested negative but subsequently tested positive and three 242 patients who had virus detected in one specimen type but not the other (NP swab but not 243 OP swab detected n=1; NP/OP swab but not N P swab detected n=1, and OP swab but 244 not NP swab detected n=1). One inpatient initially tested positive with B-ßCoV, and then 245 positive with medium viral burden SARS -CoV-2 two days later. Twelve of the thirteen 246 patients who converted from negative to positive were initially tested at the E mergency 247 Department (ED). On repeat testing performed within three days, 5 had low viral burden, 248 5 had medium viral burden, and 2 had high viral burden on the positive test. Of the ten 249 patients with low-medium viral burden, nine subsequently tested positive as inpatients 1-250 3 days later , and t he tenth tested positive at the ED 3 days later. Two patients tested 251 positive with high viral burden the next day as inpatients. The thirteenth patient tested 252 negative as an inpatient and then with high viral burden as an inpatient seven days later. 253 Of note, while most of the patients in the dataset presented with symptoms and were 254 being tested for suspected infection with SARS-CoV-2, obstetrics and gynecology (OB) 255 patients in the labor and delivery wards were being universally screened for SARS-CoV-256 2 as a pre -procedural measure to determine if personal protective equipment would be 257 required during interactions with healthcare workers. This group consisted of 102 female 258 patients with a 7% positivity rate. 259 260 Means of turn-around-times from test order to result and time -in-lab to result were 19.8 261 (13.1-26.2) hours and 11.9 (7.0-24.0) hours, respectively. The percentage of tests with 262 detected SARS-CoV-2 increased as the weeks progressed, and settled at approximately 263 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 10 50% from 3/21/2020 to 3/30/2020 ( Figure 4a). Most of the samples were from the ED 264 (n=911), followed by inpatient wards (n=492) and outpatient clinics (n=113) (Table 4), 265 and the highest positivity rate was in the E D with 50% of patients with detected SARS -266 CoV-2 (p=0.0005). There was a significant difference in the age (p=0.0005) and gender 267 (p=0.005), with lower rates of detected virus in younger patients and female patients. Only 268 7% of patients 18 years and under had detected virus. Within female patients, older 269 female patients (>55 years, n=346) tested positive with greater frequency than younger 270 female patients (< 55 years, n=438) (p=0.001), while this was not the case with male 271 patients in the same age ranges (p=0.09) (Figure 4b). This effect was diminished after 272 removing patients from the labor and delivery ward (102 patients, 7% positive) who were 273 being screened universally regardless of symptoms (p=0.03). There was no significant 274 difference in the frequency of positive tests in different race groups (p=0.385). 275 276 Lower respiratory tract specimens, including sputum, BAL, and tracheal aspirates, were 277 accepted starting 4/17/2020. As of May 15, 2020, ten sputum, 30 BAL, and 101 tracheal 278 aspirate specimens h ad been received from 115 patients, with 0%, 13%, and 23%, 279 respectively, showing detectable SARS -CoV-2. Indeterminate results were reported for 280 three tracheal aspirate samples. The mean Ct values for E gene, S gene, and I.C. targets 281 in positive LRT samples (n=27) were 27.3 (7.7-39.1), 26.7 (6.7 -38.5), and 31. 4 (27.9-282 44.7), respectively (Figure 3). Ct values were not significantly different for the E gene, S 283 gene, and I C targets between positive BAL and tracheal aspirate samples 284 (Supplementary Figure 1). The mean number of LRT samples received per day was 7 285 (range 1-25), which was significantly lower compared to the number of URT specimens 286 tested (mean 85; range 12 -176). Given the small sample size, additional statistics on 287 clinical cohort characteristics were not calculated for LRT specimens. 288 289

290 Since the end of December 2019 , when China first reported cases of the novel 291 coronavirus disease to the World Health Organization (WHO), SARS-CoV-2 has spread 292 to dozens of countries around the world, including the United States. A rapid and accurate 293 diagnosis of infectious disease is critical for managing outbreaks. Given the increasing 294 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 11 number of people infected with SARS-CoV-2 and the lack of any commercially available 295 tests, on February 4, 2020, the FDA in the U.S. opened a pathway that allowed 296 laboratories to implement laboratory-developed tests to meet this diagnostic need . The 297 CDC made their validated kits available through Integrated DNA Technologies, but the 298 kits were very limited in number and were only approved for use with specific instruments, 299 reagents, and controls. Due to a surge in demand for SARS-CoV-2 testing, issues with 300 scaling up numbers of tests per run using the CDC method, and the limited availability of 301 kits, on February 29, 2020, the FDA issued a new policy to help expedite the availability 302 and capacity of diagnostic testing in the U.S. [6] 303 304 Altona Diagnostics launched the RealStar® SARS-CoV-2 Real-Time RT-PCR Kit as soon 305 as the disease spread to Europe on February 20. Prototype kits were sent to reference 306 centers for testing and confirmation of functionality with clinical samples [11]. The 307 reagents were designed as a dual target assay and manufactured according to GMP 308 guidelines such that the read y to use kit allow ed for rapid detection of all lineage B -309 betacoronaviruses and SARS -CoV-2 specific RNA in a single reaction. Laboratories 310 could also use the reagents with a wide range of different extraction and real -time 311 thermocycler instruments, allowing for greater flexibility in implementation. The first batch 312 of reagents arrived in our laboratory within five days from ordering on Friday, March 6, 313 2020, around the same time , we received the WRCEVA RNA reference material. The 314 requirements for obtaining the WRCEVA RNA reference material were 1) that it was used 315 for diagnostics in a CLIA-certified, high complexity lab, and 2) the laboratory was planning 316 to submit an EUA application to the FDA . This RNA reference control was instrumental 317 for the timely development of the EUA test, and notably, many laboratories had difficulties 318 obtaining controls . At the same time , a well -characterized in -house collection of 319 respiratory NP swabs and sputum samples had been collected and saved in the clinical 320 microbiology laboratory, allowing for the preparation of a negative pool of samples for 321 generating the LOD dilution and accuracy samples as described in the Methods section. 322 323 Accuracy studies of NP and sputum samples in our laboratory showed excellent overall 324 agreement between the expected and obtained results for contrived clinical specimens 325 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 12 and patient samples tested by NYC-DOH . A high ly sensitive test is crucial for the 326 detection and identification of SARS-CoV-2 in individuals exhibiting signs and symptoms 327 of a respiratory infection to allow early initiation of therapy . The LODs for the N P and 328 sputum samples were 2.7 gene copies/reaction and 23.0 gene copies/reaction, 329 respectively, suggesting a slightly higher analytical sensitivity for the NP specimens 330 compared to sputum. Overall sensitivity results for this assay by specimen type has been 331 in agreement with the LODs reported in the literature by other studies [11, 12]. Of note, 332 even though the sputum samples had a lower analytical sensitivity, higher viral loads have 333 been reported in sputum specimens compared to NP swab specimens, with LRT samples 334 being the most likely specimen type to test positive for the virus in COVID-19 patients 335 [13]. Based on these favorable validation results, we decided to start routine SARS-CoV-336 2 testing with the RealStar ® SARS-CoV-2 rRT-PCR reagents on March 11, 2020. The 337 entire FDA-EUA validation, followed by a successful go-live testing day, took only four 338 days from receipt of the reagents and RNA reference control in the laboratory. 339 340 Overall during this time, we have tested 1,694 URT (40% positive) and 141 LRT (25% 341 positive) specimens, from 1,571 and 115 patients, respectively. The lower number of 342 LRT compared to URT specimens reflects hospital policy, restricting LRT testing to 343 intubated patients that n eeded clearance of isolation (two negative N P swabs plus one 344 negative LRT specimen) or patients with high suspicion for COVID -19 with repeat 345 negative testing by RT-PCR (two negative NP swabs). 346 347 In the cohort of patients tested over three weeks by our assay, positive results were seen 348 more frequently in older males compared to younger and female patients, which has been 349 supported by several studies [14, 15]. Post-menopausal women have been reported to 350 have a greater risk of hospitalization compared to non-menopausal women attributed to 351 the protective effect of estrogen [16]. In this study, older women were more likely to test 352 positive for SARS-CoV-2 compared to younger female patients. However, the difference 353 in detection rate between older and younger women was diminished after removing 354 obstetrics patients screened universally regardless of symptoms, highlighting the 355 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 13 importance of restricting comparisons of positivity rates to groups of patients subjected to 356 similar selection criteria and warranting the importance of carefully designed studies. 357 358 Among the obstetrics patients, only 7% tested positive for SARS-CoV-2, which is similar 359 to the prevalence (13.5%) obtained for women admitted at delivery at other NYPH 360 campuses [17]. We did not see any differences in the number of positive tests by race, 361 but this was early in the epidemic in New York City. The ED likely had more positive tests 362 since patients tend to be more acutely symptomatic there compared to ambulatory clinics. 363 The percentage of positive tests increased steadily and settled at around 50% three 364 weeks into the epidemic, with later testing on other platforms showing daily positivity rates 365 as high as 75-80% as the epidemic reached its peak in specific boroughs (unpublished 366 data). In this study , 13 patients tested positive after initial negative results in the E D, 367 suggesting they had sufficient symptoms to warrant inpatient admission despite negative 368 testing. This conversion may be due to increased viral burden on subsequent days post-369 infection, or due to better sampling [18]. 370 371 In summary, we described the clinical development and implementation of an FDA EUA 372 laboratory validated rRT-PCR test for SARS-CoV-2 in our academic institution, providing 373 a road map to assist others in establishing similar tests. We also described the clinical 374 and testing characteristics of the first cohort of COVID-19 patients admitted to our 375 institution during the early days of the viral outbreak in NYC. 376 377 378 379 380 381 382 383 384 385 386 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 14 ACKNOWLEDGMENTS 387 We would like to thank all the dedicated medical technologists and health care 388 professionals who performed and assisted in testing at the clinical laboratories of NYPH-389 WCM. We also thank Dr. Scott C. Weaver, World Reference Center for Emerging Viruses 390 and Arboviruses (WRCEVA), for providing us with viral RNA control material, and Altona 391 Diagnostics for their prompt supply of reagents and support. 392 393 Competing interests: None declared. 394 Ethical approval: Obtained. 395 396 397 398 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 15 399

400 1. Lorusso, A., et al., Novel coronavirus (SARS-CoV-2) epidemic: a veterinary 401 perspective. Vet Ital, 2020. 56(1): p. 5-10. 402 2. Wei, X., X. Li, and J. Cui, Evolutionary perspectives on novel coronaviruses 403 identified in pneumonia cases in China. Natl Sci Rev, 2020. 7(2): p. 239-242. 404 3. Chen, L., et al., Clinical Characteristics of Pregnant Women with Covid-19 in 405 Wuhan, China. N Engl J Med, 2020. 406 4. Chen, N., et al., Epidemiological and clinical characteristics of 99 cases of 2019 407 novel coronavirus pneumonia in Wuhan, China: a descriptive study. Lancet, 408 2020. 395(10223): p. 507-513. 409 5. Goyal, P., et al., Clinical Characteristics of Covid-19 in New York City. N Engl J 410 Med, 2020. 411 6. Policy for Diagnostics Testing in Laboratories Certified To Perform High 412 Complexity Testing Under the Clinical Laboratory Improvement Amendments 413 Prior to Emergency Use Authorization for Coronavirus Disease-2019 During the 414 Public Health Emergency; Immediately in Effect Guidance for Clinical 415 Laboratories and Food and Drug Administration Staff (FEDERAL REGISTER). 416 Accessed on March 1, 2020. 417 7. Team, R.C., R: A language and environment for statistical computing. R 418 Foundation for Statistical Computing. Vienna, Austria, 2019. 419 8. Chen, F., CLSI EP17A2E Guidance with Application to Quantitative Molecular 420 Measurement Procedures. R Function Library, 2016. 421 9. Global Initiative on Sharing All Influenza Data (GISAID) , https://www.gisaid.org/, 422 2020. Accessed on April 30, 2020. 423 10. Craney, A.R., et al., Comparison of Two High-Throughput Reverse Transcription-424 Polymerase Chain Reaction Systems for the Detection of Severe Acute 425 Respiratory Syndrome Coronavirus 2. J Clin Microbiol, 2020. 426 11. Konrad, R., et al., Rapid establishment of laboratory diagnostics for the novel 427 coronavirus SARS-CoV-2 in Bavaria, Germany, February 2020. Euro Surveill, 428 2020. 25(9). 429 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 16 12. Corman, V.M., et al., Detection of 2019 novel coronavirus (2019-nCoV) by real-430 time RT-PCR. Euro Surveill, 2020. 25(3). 431 13. Wang, W., et al., Detection of SARS -CoV-2 in Different Types of Clinical 432 Specimens. JAMA, 2020. 433 14. Tian, S., et al., Characteristics of COVID-19 infection in Beijing. J Infect, 2020. 434 80(4): p. 401-406. 435 15. Lai, C.C., et al., Asymptomatic carrier state, acute respiratory disease, and 436 pneumonia due to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-437 2): Facts and myths. J Microbiol Immunol Infect, 2020. 438 16. Ting Ding, J.Z., Tian Wang, Pengfei Cui, Zhe Chen, Jingjing Jiang, Su Zhou, Jun 439 Dai, Bo Wang, Suzhen Yuan, Wenqing Ma, Lingwei Ma, Yueguang Rong, Jiang 440 Chang, Xiaoping Miao, Xiangyi Ma, Shixuan Wang, A Multi-hospital Study in 441 Wuhan, China：Protective Effects of Non-menopause and Female Hormones on 442 SARS-CoV-2 infection. 443 https://www.medrxiv.org/content/10.1101/2020.03.26.20043943v1, 2020. 444 17. Sutton, D., et al., Universal Screening for SARS-CoV-2 in Women Admitted for 445 Delivery. N Engl J Med, 2020. 446 18. Natalie N. Kinloch, G.R., Chanson J. Brumme, Winnie Dong, Weiyan Dong,, 447 R.B.J. Tanya Lawson, Julio S.G. Montaner, Victor Leung, Marc G. Romney,, and 448 N.M. Aleksandra Stefanovic, Christopher F. Lowe, Zabrina L. Brumme, 449 Suboptimal biological sampling as a probable cause of false-negative COVID-19 450 diagnostic test results. 451 https://www.medrxiv.org/content/10.1101/2020.05.05.20091728v1.full.pdf, 2020. 452 453 454 455 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 17 FIGURES 456 457 458 459 Figure 1. Limit of detection (LOD) studies. Ct values for the LOD serial dilution study 460 using WRCEVA RNA reference material spiked in pooled negative (A) nasopharyngeal 461 (NP) specimen eluate and (B) sputum specimen eluate. Six ten -fold dilutions were 462 performed starting at 1,000,000 gene copies/reaction and ending at 1 gene copy/reaction. 463 The apparent LOD was between 1 and 10 gene copies/reaction for NP specimens and 464 between 10 and 100 gene copies/reaction for sputum specimens. 465 466 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 18 467 468 469 470 Figure 2. LOD of NP and sputum by probit analysis. Additional serial dilution stud ies 471 were performed using WRCEVA RNA reference material spiked in pooled negative (A) 472 NP specimen eluate and (B) sputum specimen eluate to determine the LOD. Five 473 replicates (A, B, C, D, E) of six ten -fold dilutions were performed starting at 1,000 gene 474 copies/reaction and ending at 0.1 gene copies/reaction for NP and five replicates of three 475 ten-fold dilutions were performed starting at 100 gene copies/react ion and ending at 1 476 gene copies/reaction for sputum. An additional five replicates were performed at 0.8, 0.6, 477 C D All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 19 0.5, 0.4, and 0.2 gene copies/reaction for NP and 80, 60, 50, 40, and 20 gene 478 copies/reaction for sputum . Probit analysis showed LOD to be (C) 2.7 gene 479 copies/reaction for NP and (D) 23.0 gene copies/reaction for sputum. 480 481 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 20 482 483 Figure 3. Distribution of Ct values for E gene, S gene, and Internal Control targets for all 484 upper respiratory tract (URT) and lower respiratory tract (LRT) specimens with detected 485 SARS-CoV-2. Mean Ct values between URT and LRT specimens were significantly 486 different for the E gene (p=0.006) and S gene (p=0.03) but not the I .C. (p=0.7), though 487 the much smaller sample size for LRT is noted. 488 489 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 21 490 491 492 Figure 4. SARS-CoV-2 results by date and distribution by gender and age. A) Positivity 493 of URT specimens tested by rRT -PCR at NYP -WCMC over the first three weeks of 494 implementation. B) Age distribution histogram s with overlays of normalized density 495 curves corresponding to age distribution (yellow) and SARS-CoV-2 positivity (red) in 496 tested patients by gender. Patients that were universally screened at labor and delivery 497 were removed from this analysis. 498 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 22 TABLES 499 500 Dilution Gene copies per reaction Number run Number detected Percent Cy5 (S gene) Mean Ct FAMTM (E gene) Mean Ct JOETM (I.C.) Mean Ct NP SPU NP SPU NP SPU NP SPU NP SPU NP SPU 1:10 106 3 3 3 3 100% 100% 14.6 13.1 15.2 14.5 30.6 33.0 1:102 105 3 3 3 3 100% 100% 18.0 16.5 18.7 17.9 29.0 30.4 1:103 104 3 3 3 3 100% 100% 20.9 19.9 21.6 21.4 28.3 30.8 1:104 103 3 3 3 3 100% 100% 24.6 23.8 25.2 25.3 29.4 29.7 1:105 102 3 26 3 26 100% 100% 27.9 31.1 28.4 31.0 30.2 29.8 1:106 10 23 3 23 0 100% 0% 32.3 ND 32..0 ND 30.8 29.7 1:107 1 3 3 0 0 0% 0% ND ND ND ND 29.8 29.8 501 Table 1. Limit of detection studies were performed for NP VTM (NP) and sputum (SPU) 502 specimen types with three replicates at each dilution. An additional twenty (NP) and 503 twenty-three (SPU) specimens were tested at the estimated LOD of 10 gene 504 copies/reaction and 100 gene copies/reaction, respectively. 505 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 23 Validation sample altona SARS-CoV2 rRT-PCR NP-VTM Sputum Positive Negative Positive Negative *Positive (contrived) 30 0 30 0 Positive (patient) 4 0 1 0 Negative (patient) 0 30 0 30 Total 64 60 31 31 *SARS-CoV-2 contrived positive samples generated with RNA control spiked into pooled negative eluate or dilutions of a high positive clinical sample. 506 Table 2. Accuracy studies for NP and sputum. Clinical evaluation of the RealStar® SARS-507 CoV-2 rRT-PCR assay using automated QIAsymphony total nucleic acid (TNA) extraction 508 followed by rRT-PCR targeting the E and S coronavirus genes on an in-house validation 509 panel consisting of patient and contrived samples for NP (124 ) and sputum (62). 510 511 512 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 24 513 Specimen Type Cy5 (S gene) Mean Ct FAM™ (E gene) Mean Ct JOE™ (I.C.) Mean Ct Number tested Correctly classified NP swab 28.8 (20.6-37.2) 28.8 (20.8-37.7) 29.6 (28.6-30.8) 34+/30- 100% Sputum 24.1 (16.2-30.0) 24.2 (17.0-29.2) 30.2 (29.4-32.3) 31+/30- 100% BAL 23.4 (16.9-29.3) 23.5 (17.1-29.4) 30.4 (28.6-32.4) 20+/20- 100% Trach asp 22.1 (10.0-43.0) 19.7 (10.1-32.1) 34.1 (29.7-42.4) 5+/5- 100% 514 Table 3. Summary table of accuracy studies for all specimen types. Mean and range of 515 Ct values are shown for positive samples. The number of positive (either contrived 516 through spiking RNA into a negative matrix or actual patient samples) and negative 517 specimens are also noted along with the percent of specimens that were correctly 518 classified as positive or negative. 519 520 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint 25 521 Feature All Patients n=1579 Positive 38% Negative 62% p-value Age (years) Overall: 53.4 (0.1-120.3) 0-18 80 19-35 295 36-55 420 56-85 656 >85 128 Overall: 57.5 (1.3- 120.3) 0-18 7% 19-35 30% 36-55 40% 56-85 47% >85 30% Overall: 51.1 (0.1- 97.5) 0-18 93% 19-35 70% 36-55 60% 56-85 53% >85 70% 0.0005* Gender F 784 M 778 Unsp. 3 F 31% M 46% Unsp. 33% F 69% M 54% Unsp. 66% 0.005* Race Asian 101 Black 171 Declined 173 Other 210 White 488 Asian 28% Black 36% Declined 40% Other 45% White 32% Asian 72% Black 64% Declined 60% Other 55% White 68% 0.385 Location Emergency 911 Inpatient 492 Outpatient 113 Emergency 50% Inpatient 18% Outpatient 35% Emergency 50% Inpatient 82% Outpatient 65% 0.0005* 522 Table 4. Summary table of patient characteristics. For race, “Declined” and “Other” 523 categories were not used when performing the Chi-squared test for significance. An 524 additional 63 tests were performed at low numbers at several other locations; these were 525 not included in the table. 526 527 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted June 8, 2020. ; https://doi.org/10.1101/2020.06.05.20109637doi: medRxiv preprint