Methods
We performed a retrospective cross-sectional study on data from patients with a SARS-CoV-2-positive diagnostic nasal-pharyngeal swab analysed by our regional reference Laboratory (Amedeo di Savoia Hospital, Turin, Italy) in March 2020, when diagnostic samples from suspected cases from the entire region where mainly collected by our laboratory.
Patients were stratified according to diagnostic Ct values into three groups: Ct≤20.0, group A; 20.0<Ct≤ 28.0, group B; Ct>28.0, group C. The 28.0 cut-off was chosen due to the 100% detection rate of SARS-CoV-2 rapid antigen in samples with a Ct≤28, previously described by our Laboratory 21. Patients with a Ct≤ 28.0 show higher viral load and likely a still replicating virus as testified by viral antigen expression21. A further stratification within this group was made according to the second cut-off of 20.0 to divide the subjects according to the amount of viral load and potential infectivity, as suggested by preliminary results of studies transfecting cell cultures by diagnostic specimens 22–24.
Swabs were processed by the 2019 Novel Coronavirus Real-Time Multiplex PCR kit (Liferiver Bio-Tech, San Diego, CA, USA), targeting three SARS-CoV-2 specific genes: RNA-dependent RNA Polymerase (RdRP), Nucleocapsid, and Envelope. For the purpose of the study, only RdRp Ct values were considered to have one uniform proxy of viral load, being RdRp the most specific gene among the three. The ABI Prism 7500 thermal cycler (Thermo Fisher Scientific, Waltham, MA, USA) was used for PCR amplification.
In March, 1995 samples resulted SARS-CoV-2-positive at our Laboratory: 1138 swabs with Ct value>28.0 and 857 below. The present study was primarily designed to assess whether there was a difference in mortality between cases stratified according to Ct value at COVID-19 diagnosis. Therefore, to detect as significant an estimated difference in mortality of at least 0.2 [], with a two-sided confidence level of 0.05, a power of 95% and a ratio between Ct values higher and lower than 28.0 of 1.3, an overall sample size of 225 would be required.
Patients were randomly sampled from the frame represented by the 1995 SARS-CoV-2-positive swabs through probability sampling (random lottery extraction). The sampled individuals were reached in August-September 2020 for a telephone survey addressing COVID-19 related clinical and demographic characteristics both of the interviewed and of his/her household contacts (an English translation of the survey is shown in Suppl.Fig.1). The surveyed data were crosschecked and completed by data extrapolated by the Piedmont platform (RUPCOVID), an on-line regional database built up for SARS-CoV-2 contact tracing, notification (swab results and dates) and clinical data collection (demographics, signs and symptoms at onset and at diagnostic swab, date of symptoms onset, comorbidities).
Diseases severity was classified according to a 6-degree scale as follows: no hospital admission, hospitalisation without oxygen support, hospitalisation with support from low-flow wall oxygen to reservoir mask, hospitalisation requiring continuous positive airways pressure (CPAP) support, hospitalisation with intubation, death. Signs and symptoms were clustered according to four main systems: fever, asthenia, malaise and arthromyalgia as inflammatory systemic involvement; headache, olfactory and gustatory dysfunction as neurological involvement; nausea, vomiting and diarrhoea as gastrointestinal involvement; dyspnoea, runny nose, cough and pharyngitis as respiratory involvement.
Patients not consenting to the telephone survey were discarded and their data not collected from the RUPCOVID. Anonymized data were used for deceased subjects. The study was approved by the Regional Department for Infectious Diseases and Emergency (DIRMEI, Torino, Italy).
Data were analysed through nonparametric tests (Mann-Whitney, Chi-square for trend, Kruskal-wallis and Fisher exact tests). Variables with relevant biological significance or showing univariate p≤0.10 were included in the multiple linear or ordinal logistic regressions (entry method). Categorical variables are presented as absolute number (proportion) while continuous variables as median (interquartile range). Data analysis was performed through SPSS 25.0 (IBM stat.).