An update on COVID testing

Air Canada and McMaster HealthLabs COVID testing trial - new results published

Six weeks ago, I reanalysed the preliminary results from Canada’s airline passenger COVID testing trial in this post. International arriving passengers are being tested at Toronto airport and again at 7 days and 14 days post arrival. They are trying to quantify the effectiveness of airport testing alone, or at least trying to get the current 14 day quarantine period shortened.

I took Air Canada to task for focusing on the fact that 80% of people testing positive did so at the airport, seeming to imply that airport testing alone would identify 80% of positive cases. After correcting for the misuse of statistics, I showed that their results indicated that airport testing alone would only catch about 60% of cases.

Yesterday, the testing partner McMaster HealthLabs published more results from the study. Let’s take a look at what they show.

The results

Interestingly, the new results only cover people who had signed up by October 2nd, presumably to make sure that there had been time to get back test results after 7 and 14 days and analyse the findings. That corrects one of the issues I had with the previous set of statistics.

It also means that the sample still only includes travellers in September. That predates the second wave increase in cases in Europe and the USA, the main countries from which these passengers were arriving. The overall 1% testing positive would probably be at least 5 times higher now based on the increase in cases since then.

Helpfully, they published the number of people taking each of the tests and the number testing positive. Those figures are as follows:

Test	No. tested	Positive tests	% Positive
Airport	8,644	61	0.7%
Day 7	6,620	23	0.3%
Day 14	5,517	5	0.1%

McMaster HealthLabs conclude from these statistics that “An arrival PCR test would detect about 70% of positives”. I assume the calculation they are doing here is 61 / (61 + 23 + 5) = 68.5%, which they have rounded up.

Numbers starting with a 7 are so much nicer than those starting with a 6, don’t you think?

Correcting the statistics, again

Whilst “about 70%” is a better statistic than the highly misleading 80% headline from last month’s press release, I still don’t think it is quite right.

Note the drop off in numbers tested at day 7 and day 14 compared to those tested at the airport. Even allowing for not testing anyone who has already tested positive at an earlier stage, there is clearly quite a drop off in participation. Using the figures the way they have done implies that the 2,000 people who didn’t take the day 7 test would have tested negative. They have no reason to believe that.

Instead, I will make the more neutral assumption that the untested people would have tested positive in the same proportion as those that did take the test. That would give me the following figures for people who tested positive, or would have been expected to, at each stage.

Test	No. tested	Positive tests	% Positive
Airport	8,644	61	0.7%
Day 7	8,583	30	0.3%
Day 14	8,553	8	0.1%

That gives me a total of 99 expected positive tests for a sample of 8,644 passengers or 1.1%. Of the positive passengers, 61 would have been caught at the airport, which is 61.6% (61 / 99).

That is very close to my earlier 60% estimate and not “about 70%”. If you also allow for the fact that the test accuracy is not perfect, 60% still seems like a good estimate to me for the effectiveness of an airport arrival test.

Can pre-departure or airport only testing replace quarantine requirements?

The main reason airlines and airports are pushing for the introduction of testing is because they believe it can eliminate quarantine requirements, which is one of the main things holding back a resumption of air travel.

If we assume they got their way and that led to international travel getting back to 50% of 2019 levels, a country like Canada would be receiving about 100,000 international arrivals a week.

Based on the current prevalence rate which is at least five times higher than during September, that would mean that 5,500 infected people would be arriving on aircraft each week. If mandatory pre-departure and/or airport testing can catch 60% of them, that still leaves about 2,200 undetected imported cases a week. By way of comparison, last week Canada reported about 35,000 new COVID cases, so that’s a material increase in risk.

Of course quarantines are not perfect either, but that is an argument for adding mandatory pre-departure testing, not dropping quarantine requirements.

Can quarantine periods be reduced?

The trial provides better evidence for reducing quarantine periods to about a week. Tests at the airport and at 7 days would catch 92% of cases (91 / 99). Maybe governments could be persuaded that 90% is good enough, but a reduction in quarantine to 7 days isn’t going to kickstart tourism or business travel.

An alternative idea

Something that seems worth considering to me is relocating the quarantine period from destination to origin. If the rule was that you need to self-isolate for a week or ten days prior to taking a flight and get a negative test just before departure, then that might be a more palatable alternative for travellers than needing to quarantine away from home. Of course, that would require countries to trust each other and co-ordinate their policies. It would also need someone to trial such a policy to prove its effectiveness.

However there has been little sign of countries working together so far and I don’t see much chance of that changing. The new testing trials that are being done, such as the one just launched by British Airways, American Airlines and oneworld yesterday are still trying to demonstrate that one or two tests, taken pre-departure and/or on arrival at the airport are sufficient. It seems to me that has already been disproven by the Canada trial and other evidence about test effectiveness during the incubation period. The only hope they have is that the LAMP test being used by oneworld for the tests after arrival is better at picking up cases soon after infection than the PCR test used in the Canadian trial. Nothing I have read suggests that is likely to be true.

The trial also seems misconceived in that the third test is to be taken only three days after arrival. Getting a negative test after three days is not reliable evidence that the passenger is not infected. We know from the Canada results that a clear test even at 7 days will miss 10% of cases that will be picked up by Day 14. The best that can come out of the oneworld trial is that the Day 3 test is almost as good as a Day 7 test. Ironically, oneworld should be hoping to catch a large number of additional people in the Day 3 test. That would support an argument for reducing quarantine to 3 days with a negative test at the end.

If only a few extra cases are detected at Day 3, which is what I imagine oneworld are hoping for, all that will tell you is that Day 3 is too early to pick up all the cases you missed at the airport. To disprove that, oneworld would need to add a fourth or fifth test at Day 7 or 14 which showed very few additional positive cases.

My proposal for modifications to the oneworld trial

What I think oneworld need to do is to move the timing of the third test to seven days after arrival. That will provide data that can be compared with the Canada results. If their data shows that the ratio of additional positive tests at Day 7 to positive tests is lower than found in the Canada trial, that would be good evidence that their LAMP test is better at picking up early infections than the one used in Canada.

Secondly, they need to try and get a decent sub-sample of passengers who certify that they have self-isolated for seven days prior to taking their pre-departure test. If they can prove that for this group, the number of infections caught at Day 7 which were missed at the airport is much lower, then maybe they will have evidence to persuade governments to allow a “quarantine at home” alternative for the outbound leg of the journey.

Maybe it is all just politics and PR

Perhaps I am making the mistake of trying to answer the question of whether pre-departure and airport testing is genuinely capable of adequately reducing the risk of importing cases, such that quarantines can be safely removed and that people can be encouraged to travel again. The evidence we have seems quite clear that the answer to that is no. Maybe better tests and additional evidence from more trials can change that, but it seems quite unlikely to me.

The industry argues that countries like Germany have been successfully using testing as an alternative to quarantine. When Germany’s infection rates were way below other European countries, that may have been a decent argument. But Germany has not escaped the second wave and infection rates are now not far below those of other European countries.

Of course, the objective of these testing trials may in fact be just to generate nice sounding headlines. The statistics may not stand up to proper scrutiny but they can provide political cover for relaxing the restrictions to save travel sector jobs and allow air travel to resume.

At some point, when the prevalence of the virus has dropped back that will be the right “balance of costs and risks” judgement for politicians to make.

I just don’t think that is likely to happen before Spring 2021 at the earliest. With airport testing only 60% effective, countries need prevalence to drop a long way before the risk of importing cases becomes low enough to want international travel to get going again.

Mandatory pre-departure and/or arrival testing may well be coming, as it would obviously reduce risk, since quarantines are far from perfect. But however much airlines and airports push for the complete elimination of quarantine, the evidence so far doesn’t support that. I think public health experts know that compliance with quarantine requirements is not great, but they do know that it acts as an effective deterrent to travel. Cases can’t be imported if nobody is travelling.

All the lobbying and pressure from the travel industry may just result in mandatory testing being added, but without eliminating the quarantine requirements. That will hurt not help travel businesses in the near term, although it might help accelerate the day when the case rates drop to levels where neither will be needed.

In the meantime, don’t hold your breath and expect testing to come to the rescue any time soon.