Jon Deeks Profile picture
Biostatistician and trialist trying to work out the best ways of evaluating medical tests.
25 Jun 20
Antibody tests for identification of current and past infection with SARS‐CoV‐2 - Deeks, JJ - 2020 | Cochrane Library cochranelibrary.com/cdsr/doi/10.10…

Living Cochrane Review of Antibody Tests – 1st version now published. Twitter summary to save you reading the 300+ pages
#1 Search date of 27 April 2020. Scanned 10,965 COVID-19 publications and pre-prints. Included 54 studies of test accuracy of COVID-19 antibody tests. Some studies looked at multiple tests, so total of 89 test evaluations including 15,976 samples (8,256 from COVID-19 patients)
#2 Most research studies done in China on tests developed and manufactured in China. Data found for 27 commercial tests and 25 in-house assays. 61% were lab assays (requiring venous blood), 39% with point-of-care assays (possible finger-prick blood)
Read 22 tweets
28 May 20
TRACK-AND-TRACE: MANY PCR-ves WILL HAVE COVID-19

Government advice (gov.uk/guidance/nhs-t…)

“If your test is negative, you and other household members no longer need to self-isolate”

BUT chances of COVID could be 19% in this group
@petermbenglish @drhelmac @PHE_UK @drhelmac
DATA (*more needed*)

Of staff in UK hospital 43% with week off for COVID symptoms were RT-PCR+ve.

Sensitivity of RT-PCR - no clear data - stated to be as low as 70%.

Investigate worst case 43%, compare with 24% best case
Sens: 70% to 85%
Spec: 98% (makes little difference).
TECHNICAL STUFF #1 – NEGATIVE LIKELIHOOD RATIOS (LR-)

LR- = (100-sens)/spec
Read 7 tweets
16 May 20
21 WAYS TO SPIN RESULTS FROM A COVID DIAGNOSTIC TEST ACCURACY (morning thoughts whilst walking the dog - please add others I've missed). I'm putting this out to help us critique the "number theatre" being thrown at us.
#diagnostics #COVID19 @deb_cohen @MHRAgovuk @drhelmac
1/22
#1 Use a small sample size (and never report how big it isn’t). Use phrases like"we detected all cases" the test is "100% accurate" without mentioning it is only from 29 samples.
2/22
#2 Talk about the total sample size when interpreting results about sensitivity (especially if you have tested 5000 blood donors to estimate specificity and only 29 to estimate sensitivity)
3/22
Read 23 tweets
14 May 20
1/9 Where’s the evidence for the ROCHE test claims?

All I can find is the IFU document. fda.gov/media/137605/d…

@deb_cohen @phe_uk @MHRAgovuk @sarahboseley I'll summarise what I have found. Please share other studies and results if you can find them. We've searched hard.
2/9 Study design

“A total of 204 samples from 69 symptomatic patients with a PCR confirmed SARS‑CoV‑2 infection were tested with the Elecsys Anti‑SARS‑CoV‑2 assay. 1 or more consecutive specimens from these patients were collected after PCR confirmation at various time points.”
3/9 (that's all the detail available - standard two sentences of detail - STARD reporting guidelines clearly not applied).
Read 9 tweets