University of Michigan - Ann Arbor

"From January to May 2020, virology laboratories at Charité and Labor Berlin screened 77,996 patients for SARS-CoV-2 infection. Of these, 3,303 (4.2%) had at least one positive result. Initial testing exclusively used Roche LightCycler 480 II (LC480) PCR instruments. From mid-March, Roche cobas 6800 and 8800 (cobas) instruments were introduced to increase testing capacity. Positive PCR results were obtained in 1382 of 26,453 (5.2%) tests on LC480 and 2216 of 67,548 (3.3%) tests on cobas instruments. The per-instrument totals sum to more than the patient total due to some patients being tested and testing positive multiple times on both systems."

Results for An analysis of SARS-CoV-2 viral load by patient age

Terry C Jones, Barbara Mühlemann, Talitha Veith, Guido Biele, Marta Zuchowski, Jörg Hoffmann, Angela Stein, Anke Edelmann, Victor Max Corman, Christian Drosten
medRxiv 2020.06.08.20125484; doi: https://doi.org/10.1101/2020.06.08.20125484

https://www.medrxiv.org/content/10.1101/2020.06.08.20125484v1

An analysis of SARS-CoV-2 viral load by patient age

"To look for relationships between viral load and age, we took two approaches: 1) testing for statistical differences between the viral loads in aggregated age categories, and 2) treating age as a continuous variable and using a gamma regression to predict viral load. In the age categorization we made comparisons between three groups of patients: a) 0-9 years versus 10-99 years, b) 0-9 years versus 19-99 years, and c) 0-19 years versus 20-99 years. Age category viral loads were compared via the Mann-Whitney rank test and Welch’s t-test. The categories were also examined via a Bayesian analysis using gamma mixture models to account for the multi-modal nature of the viral load data. The Bayesian analysis used a mixture of three gamma distributions and accounted for variations between age groups by estimating age-group specific component weights. To provide a fine-grained overview of the data, log10 viral loads for an overall 10-year age bracket breakdown, with number and percentage of RT-PCR positive patients are shown in Table 2, while Figures 4 and 5 show the distribution of
viral load values in these groups."

Statistical Approach for An analysis for SARS-CoV-2 viral load by patient age

'LC480Count’, in blue, shows the number of tests performed using the Roche LightCycler II (LC480) machine, ‘T2Count’, in orange, shows tests performed on the cobas machines over time. Six day moving average.

Figure 1: Utilization of LC480 and cobas test systems over the study period for An analysis of SARS-CoV-2 viral load by patient age

"Mean log10 viral loads in the healthy and the pre-existing condition categories are 5.642 and 4.408, respectively. Parametric (Welch’s t-test) and non-parametric (Mann-Whitney rank test) find a significant difference in viral loads between children that are healthy and those with a pre-existing condition (p-values 0.027 and 0.006, respectively), when testing the combined viral loads from both PCR systems. When only considering viral loads measured by cobas instruments, no significant difference is detected."

Figure 2: Differences in viral load in an exemplary group of patients aged 0-11 years with and
without a pre-existing condition.

Orange data points indicate tests done on LC480 machines, black on cobas. Community/cluster testing includes children tested at walk-in clinics, as well as those tested due to having contact with a SARS-CoV-2 positive individual, asymptomatic cases, and family clusters. ‘Other clinical centres’ includes children tested in hospitals. The dotted line denotes a log10 viral load of 5.4, corresponding to 250,000 viral copies. 

Figure 3: Viral loads over time in children from A) 0-9 and B) 0-19 year olds.

The base 10 logarithm of viral load is estimated from the real-time PCR Ct value. Counts of people in each age category are given in parentheses in the x-axis labels. A) LC480, B) cobas.

Figure 5: Viral load by patient 10-year age strata. 

Points are observed log 10 viral load, lines indicate expected log 10 viral load from the regression model. The shaded area shows a 95% credible interval. X-axis age range is approximately 0 to 100 years (exact range obscured for data privacy).

Figure 6: Conditional effect of age from a Bayesian gamma regression predicting viral load from
age, while adjusting for type of PCR system (LC480 or cobas).

Learn Before

Related