That's an awful lot of words with very little meat. The most significant passage seems to be this:
So I took a look at Drosten's paper:
https://zoonosen.charite.de/fileadm...s-of-SARS-CoV-2-viral-load-by-patient-age.pdf
They divide the population tested into ten age groups, ten year periods from 1-10 to 91-100. They do an alternative analysis of kindergarten, grade school, high school, university, adult, mature. The two youngest age groups have a much lower positive rate than the other age groups, which they ascribe to children being more likely to be asymptomatic, so more positive cases remain unconfirmed.
(As an aside, they also note that the children who had been hospitalized had lower viral loads than children who tested positive in more general testing regimes. They account for this by noting that hospitalized children will be further beyond initial symptom appearance, when viral loads decline. This supports the point made in my previous post, that infectivity declines rapidly a few days after symptoms appear).
The authors then measure viral loads, where it gets really messy. The two youngest groups have the lowest viral loads, less than half of any of the other age groups, except one, which is still about 50% higher. The same pattern is seen with the other classification, except that the grade school group actually has the highest viral load. But kindergarten and high school are less than half of the others. This is consistent with the quote that the children have 67-85% less loads. In fact, the mean viral load of the 1-10 group is about 70% lower than that of the lowest (but one) of the older groups, and almost 90% lower than the highest of the older groups.
The problem is that there is considerable variability, particularly for the two youngest age groups (and the oldest age group), which have the smallest sample sizes (5-10 times less than the others). This results in a much larger standard error, which obscures the differences between these two youngest ages and the other ages. Looking at the data, if the sample size for these two groups were comparable to that for the other groups, so the standard errors were the same as for the latter, with the same means, the differences definitely would be significant.
To complicate matters further, the authors show that the viral loads are not distributed normally, neither for the entire dataset, nor for each age group. Looking at the data, it appears that as the groups get older, a more normal distribution is approached, but it's never completely normal, and in any case, some of this effect may be because the sample size is generally greater for the older groups.
Why are the viral loads not normally distributed? The authors don't comment on this, but from the infectivity curve I referred to in a Nature Med paper, one would expect that since viral loads start declining after symptoms appear, and people are generally tested only after symptoms appear, viral loads would have to follow a non-normal distribution. They're being measured not along a symmetric curve, but only over the declining half of such a curve (which isn't symmetric, for that matter). Now some people might be tested because they were in close contact with someone with symptoms or who has tested positive, in which case they might be on the ascending portion of the curve. But that's likely to be only a few days, whereas the descending portion goes on for much longer.
Anyway, because of the lack of normality, the authors chose several statistical tests that don't depend on this. They performed pairwise comparisons between each of the groups, and found no significant differences for the ten age groups. Using the other group classification, they found a significant difference between the kindergarten group and the oldest group, but no other significant differences.
I'm not qualified to comment on the statistics, except to repeat, the relatively small sample size of the two youngest groups is a serious problem, but this is clearly where the criticisms of Drosten's study are coming from. The very big differences in means strongly suggest that children do have significantly lower viral loads, but the statistics Drosten uses can't establish this. Whether other statisticians can, just from the data presented, I don't know, but clearly an analysis of more positive children could resolve this.
Hopefully. However, given, as I mentioned before, that viral loads depend critical on when in relationship to symptom emergence the person is tested, the timing of testing is critical. People who get tested because of recent exposure to the virus are likely to have higher loads, whereas people who get tested only several days after symptoms appear will have lower loads. . Drosten doesn't address this, but his study seems to avoid this for the most part by testing people without regard to symptoms, In other words, there is an approximately random distribution of people with regard to where they are relative to symptom appearance. But as in any large-scale testing regime that isn't truly random, we can't rule out that some people were tested because they had symptoms, or because they were in contact with people who were or might be postive. These people could skew lower or higher in viral loads.
Drosten also comments that if infected children are more often asymptomatic (though this is a common assumption, I don't think there are any data that actually support it), then they will be less likely to spread the virus by coughing. Also, because they are physically smaller than adults, they will spread the virus less distance through the air by coughing or breathing. But against those arguments, he also points out that children are more likely to engage in frequent, multiple interactions with others.
Last edited:
Facebook
Twitter
Instagram