Read on Twitter
Science India study + Munro comments (THREAD)
I recently saw @apsmunro wrote a thread on that huge-N India contact-tracing study in Science.
That’s a GOOD thing.
Post-publication scrutiny is part of science, and is especially important for literature that impacts policy. 1/ https://twitter.com/apsmunro/status/1311616478844780544
I recently saw @apsmunro wrote a thread on that huge-N India contact-tracing study in Science.
That’s a GOOD thing.
Post-publication scrutiny is part of science, and is especially important for literature that impacts policy. 1/ https://twitter.com/apsmunro/status/1311616478844780544
Some of my followers seemed concerned, so I thought I’d make a few comments.
Munro raises the important point of uncertainty in determining index cases. (The “index” transmits COVID; the “secondary case” or “positive contact” catches it.)
But we should tread carefully here. 2/
Munro raises the important point of uncertainty in determining index cases. (The “index” transmits COVID; the “secondary case” or “positive contact” catches it.)
But we should tread carefully here. 2/
To illustrate his point, Munro cites the below ADC BMJ study of 107 pediatric indexes in South Korea.
This study excluded 40 of 41 positive contacts on grounds of their being “assessed to have the same exposure as the paediatric COVID-19 index cases” 3/ https://adc.bmj.com/content/early/2020/08/06/archdischild-2020-319910
This study excluded 40 of 41 positive contacts on grounds of their being “assessed to have the same exposure as the paediatric COVID-19 index cases” 3/ https://adc.bmj.com/content/early/2020/08/06/archdischild-2020-319910
The study gives no details on how this assessment worked, but the only case that passed their standards was one in which an infected child index FLEW IN FROM THE UK.
More importantly, they ONLY excluded POSITIVE contacts on grounds of common exposure, NOT NEGATIVE contacts. 4/
More importantly, they ONLY excluded POSITIVE contacts on grounds of common exposure, NOT NEGATIVE contacts. 4/
Excluding positive contacts but not negative contacts for potential common exposure makes no sense. This would absurdly underestimate the secondary attack rate (SAR).
It’s notable that they don’t use these methods to compute the SAR for any other age group for comparison. 5/
It’s notable that they don’t use these methods to compute the SAR for any other age group for comparison. 5/
They claim the below study used the same methodology to get a 7.6% SAR for adults... but that’s false.
The below study didn’t even consider excluding any pos contacts, and it explicitly says it didn’t examine potential alternative transmission routes. 6/
https://ophrp.org/journal/view.php?doi=10.24171/j.phrp.2020.11.2.04
The below study didn’t even consider excluding any pos contacts, and it explicitly says it didn’t examine potential alternative transmission routes. 6/
https://ophrp.org/journal/view.php?doi=10.24171/j.phrp.2020.11.2.04
Munro emphasises the 107-index ADC BMJ study’s claim that it used the same data as the below SK study, but the former found 107 0-18-yr-old indexes from 20 Jan to 6 Apr. The below found 153 0-19-yr-old indexes from 20 Jan to 27 Mar.
46+ 19-yr-olds? 7/ https://wwwnc.cdc.gov/eid/article/26/10/20-1315_article
46+ 19-yr-olds? 7/ https://wwwnc.cdc.gov/eid/article/26/10/20-1315_article
Also, the Results section’s 3rd sentence has a typo: 248 household contacts for 107 indexes is 2.3 hh cont/ind, not 4.3.
But typos are fixable.
You’d have to redo the entire study to fix its SAR methodology, and SAR was its whole point.
So.. who would publish such a study? 8/
But typos are fixable.
You’d have to redo the entire study to fix its SAR methodology, and SAR was its whole point.
So.. who would publish such a study? 8/
Evidently, the RCPCH would: ADC BMJ is their journal.
In fact, @ADC_BMJ’s 3 most recent tweets are about the RCPCH, including a podcast of RCPCH president Russell Viner with the chief editor.
The same Viner who wants to reduce COVID testing for kids. 9/ https://www.theguardian.com/society/2020/sep/27/too-many-children-tested-for-covid-leading-study-schools
In fact, @ADC_BMJ’s 3 most recent tweets are about the RCPCH, including a podcast of RCPCH president Russell Viner with the chief editor.
The same Viner who wants to reduce COVID testing for kids. 9/ https://www.theguardian.com/society/2020/sep/27/too-many-children-tested-for-covid-leading-study-schools
Also the Viner who lobbied against masks in schools on the BBC in late Aug, + claimed mask use might “spread the virus more,”
https://twitter.com/BBCNewsnight/status/1298021228146495492?s=20
and who lead-authored a highly problematic study on school closures in April (see my thread
).
But I’m getting off topic. 10/ https://twitter.com/globalhlthtwit/status/1310150135268155392
https://twitter.com/BBCNewsnight/status/1298021228146495492?s=20
and who lead-authored a highly problematic study on school closures in April (see my thread
).But I’m getting off topic. 10/ https://twitter.com/globalhlthtwit/status/1310150135268155392
So...
a) Identifying the index with 100% certainty is hard (short of viral analysis).
b) Munro’s assertion that the large-N SK study and huge-N India study “suffer the same massive source of bias” and have “massive bias for children” is a useful hypothesis, but unproven. 11/
a) Identifying the index with 100% certainty is hard (short of viral analysis).
b) Munro’s assertion that the large-N SK study and huge-N India study “suffer the same massive source of bias” and have “massive bias for children” is a useful hypothesis, but unproven. 11/
Okay, back to the Science Magazine India study...
https://science.sciencemag.org/content/early/2020/09/29/science.abd7672
(I also recommend @apoorva_nyc’s article and thread.) 12/ https://twitter.com/apoorva_nyc/status/1311309459298365447
https://science.sciencemag.org/content/early/2020/09/29/science.abd7672
(I also recommend @apoorva_nyc’s article and thread
.) 12/ https://twitter.com/apoorva_nyc/status/1311309459298365447
Graphs found same-age transmission rate for kids was much higher than that for 20-39-yr-olds or 40-64-yr-olds, but this decreased as a function of age for children.
Munro’s apparent take is that authors misidentified indexes for the vast majority of child secondary cases. 13/
found same-age transmission rate for kids was much higher than that for 20-39-yr-olds or 40-64-yr-olds, but this decreased as a function of age for children.Munro’s apparent take is that authors misidentified indexes for the vast majority of child secondary cases. 13/
Munro says “Children (particularly young children) do not travel alone, especially during lock down in a pandemic,” and claims the lower same-age transmission for older kids is due to greater independence/mobility, leading to fewer misidentified accompanying-adult indexes. 14/ https://twitter.com/apsmunro/status/1311616500130820099
I have 2 comments:
1) One could alternatively interpret the age-related decrease in same-age transmission rate for kids as due to age-correlated adherence to hygiene and social distancing during socialising.
2) Munro risks projecting British culture onto Indian here. 15/
1) One could alternatively interpret the age-related decrease in same-age transmission rate for kids as due to age-correlated adherence to hygiene and social distancing during socialising.
2) Munro risks projecting British culture onto Indian here. 15/
Note I only say “risk.” I’m totally ignorant on child socialisation in India.
But in my own 80s US childhood, most non-school child interaction occurred sans adult. Even as 3-yr-olds, we’d knock on neighbour doors or find kids in neighbouring yards (+no walled gardens). 16/
But in my own 80s US childhood, most non-school child interaction occurred sans adult. Even as 3-yr-olds, we’d knock on neighbour doors or find kids in neighbouring yards (+no walled gardens). 16/
I’ve read that school in India starts in early July, that schools closed this summer, but that much of lockdown was eased in June and especially July.
It’s not inconceivable that with lockdown eased, working parents with no child care let their kids go play with friends. 17/
It’s not inconceivable that with lockdown eased, working parents with no child care let their kids go play with friends. 17/
I suspect for *any* region, the sorts of graphs in tweet 13 will be highly culture dependent: extent of horizontal v vertical age-group interaction, social patterns as impacted by locally implemented pandemic measures, typical living arrangements and childcare solutions... 18/
Eg, maybe the blip of additional transmission from 5-9-yr-olds to 0-4-yr-olds is due to young kids toddling after their next-oldest siblings during play time, before graduating to playing more exclusively with their own age group.
(But that might just be *me* projecting.) 19/
(But that might just be *me* projecting.) 19/
Then there are Munro’s assertions that
a) ~50% of children are asymptomatic,
b) symptomatic ppl seem “much more infectious” than asymptomatic ppl,
c) a+b make results from the study “not generalisable,” and
d) “in contact tracing, small and detailed beats big and dirty.” 20/
a) ~50% of children are asymptomatic,
b) symptomatic ppl seem “much more infectious” than asymptomatic ppl,
c) a+b make results from the study “not generalisable,” and
d) “in contact tracing, small and detailed beats big and dirty.” 20/
My responses:
a) Various small-N pediatric screening studies found ~45% of kids asymptomatic AT TIME OF POSITIVE PCR TEST.
That’s entirely different to asking whether they ever developed symptoms. Even many infected adults are PCR-detectable before symptom development. 21/
a) Various small-N pediatric screening studies found ~45% of kids asymptomatic AT TIME OF POSITIVE PCR TEST.
That’s entirely different to asking whether they ever developed symptoms. Even many infected adults are PCR-detectable before symptom development. 21/
b) There are conflicting results and limited consensus on relative prevalence of pre-, a-, and symptomatic transmission. Munro says “absolutely,” and cites 1 study (which omits the number of asym indexes) but there are many studies. Like this sys rev: 22/
https://www.acpjournals.org/doi/10.7326/M20-3012
https://www.acpjournals.org/doi/10.7326/M20-3012
c) Barring substantial virus mutations between India and the UK, children should exhibit similar degrees of asymptomaticity in both countries. So I just don’t know what his tweet 3/10 even means by “not generalisable.”
Maybe I’m misinterpreting him? 23/ https://twitter.com/apsmunro/status/1311616487954747393
Maybe I’m misinterpreting him? 23/ https://twitter.com/apsmunro/status/1311616487954747393
d) Heterogeneity of transmission (see below) makes many studies too small-N to capture enough transmission events. I’ve seen Munro recently cite an N=6 (3 kids + 3 adults, mostly in mutual contact) Ireland study with no at-school transmission (and limited at-school time). 24/ https://twitter.com/zeynep/status/1311329981419728898
This study shouldn’t cause panic about child transmission. There’s still a lower % of infected kids than adults, and pediatric COVID mortality is extremely low.
But neither should we claim, as Munro, that this study “doesn’t get us closer.”
It still has more to teach us.
/ENDS
But neither should we claim, as Munro, that this study “doesn’t get us closer.”
It still has more to teach us.
/ENDS
