What is the method of the second version

Statistical tests in the second version of the Drosten study

Christian Drosten's team has uploaded a second version of the preprint publication on the viral load of SARS-CoV-2 in children and adults to the Charité server (see primary source). In the past few days, the first version of this preliminary study triggered a wave of media coverage. Some of the reports also took up the technical criticism of some statisticians who had expressed concerns about the statistical testing methods used. Christian Drosten's working group then revised the analysis. Two statements from statisticians place the methodological updates in the new version of the publication.

A fact sheet from the SMC on the background to statistical testing in general and in relation to preprint publicationdo you think ...? Find here.

Overview

     

  • Jun.-Prof. Dr. Dominik Liebl, junior professor at the Institute for Finance and Statistics, University of Bonn
  •  

  • Prof. Dr. Kevin McConway, Professor Emeritus of Applied Statistics, The Open University, Milton Keynes, United Kingdom
    and Prof. Sir David Spiegelhalter, Chair of the Winton Center for Risk and Evidence Communication, University of Cambridge, United Kingdom

  •  

Statements

Jun.-Prof. Dr. Dominik Liebl

Junior professor at the Institute for Finance and Statistics, University of Bonn

“The methodical part of the statistical analysis in the new version of the preprint Jones et al. (2020) has been significantly improved from my point of view. In addition to the methodological improvement, the biggest difference to the first version is that two separate analyzes are now carried out, one for each PCR test type LC480 and cobas. The basic message of the first version of the preprint remains essentially the same. In the data for the cobas PCR test, the authors found statistically significant differences in viral load between children and adults. The results are unclear in the data for the LC480 PCR test, as no statistically significant differences were found. There are some restrictions, but these are clearly stated. For example, the consideration of further illness data would have been desirable - however, these relevant data could not (yet) be collected. The new version of the preprint will certainly continue to be discussed in academia, and this is a good thing. The authors have further improved their preprint with the help of criticism from academia. So the #TeamScience works and this is perhaps the only good news in a story that shouldn't have been. "

Prof. Dr. Kevin McConway

Professor Emeritus of Applied Statistics, The Open University, Milton Keynes, United Kingdom

Prof. Sir David Spiegelhalter

Chairman of the Winton Center for Risk and Evidence Communication, University of Cambridge, United Kingdom

“We are excited about the new version of the preprint on viral load by Jones et al. We believe that this revision has been greatly improved, and the authors have addressed the three main statistical problems that we addressed in our article on the first version have [1]. First, they also make it clearer that there are inevitable problems with the data that arises from being from tests done in clinical practice rather than being collected for research purposes. Second, they explain why it is still important and useful to present the data and analysis because better data is not available. And third, they make it clearer that this is a preprint publication, that the research has not yet been evaluated by other scientists or clinicians, and that it should not yet be used to guide clinical practice. "

"We welcome this greatly improved revision, and while we may provide additional comments to the authors through the usual preprint annotation channels, we do not intend to say otherwise about the revised version."

“Originally, our statement contained some concerns about the last sentence in the abstract, which did not coincide with other statements in this new version of the preprint. After reading this version of our statement, the authors agreed that the sentence did not express what they intended and have already made a small change to the sentence to match its intended meaning and remove the apparent contradiction(the uploaded version has already been changed, editor's note). We think this is a good example of how science and communication and collaboration between scientists should work. "

Information on possible conflicts of interest

All: No information received.

Primary source

Jones T et al. (2020): An analysis of SARS-CoV-2 viral load by patient age. Preprint publication on the Charité - Universitätsmedizin Berlin server.

References cited by the experts

[1] McConway K et al. (25.05.2020): Is SARS-CoV-2 viral load lower in young children than adults? Jones et al provide evidence that it is (in spite of their claims to the contrary). Medium.