Are there more false positive corona tests?

Returning travelers from risk areas, regionally even everyone who wants to: The number of corona tests has recently been greatly expanded. Is there a threat of a massive increase in the number of false positives?

The suspicion has been circulating for some time: Could it be that, with the current test strategy, many corona evidence is false positive, meaning that those affected are not infected at all - with far-reaching consequences? Christian Drosten, whose laboratory is responsible for expert advice on coronaviruses, does not see this problem. In the calculations, decisive factors such as multiple tests are not taken into account. An overview:

It's all about this

There are several ways to determine whether you are infected with Sars-CoV-2. The most common variant currently is the so-called PCR tests. A smear is taken from the throat and / or nose, which is examined in the laboratory for genetic material of the viruses. Such tests are fairly accurate, but in very rare cases - this is fairly undisputed among experts - they can also be wrong. As a result, in individual cases a non-infected person is diagnosed as being infected. The question is how do these so-called false positives work.

What are the fears?

The corona infection rate in the population in Germany is currently comparatively low. At the same time, the virus is tested quite widely, even in people who have no symptoms and for whom there is no specific suspicion - for example because of sick contacts. This unprovoked testing means that the proportion of those who are actually infected is small among all those tested.

The thesis is now that the actually very small proportion of false positive tests becomes significant and these false results are numerically much more significant. This is dealt with, for example, in a thesis paper by several German health experts. If the fear were correct, the number of new infections recorded might be too high. Many people would be sent into quarantine unnecessarily and wrongly assume that they have Corona behind them.

A calculation example

Dagmar Lühmann, Vice-Chair of the Evidence-Based Medicine Network (EBM Network), has published a paper to draw attention to the possible difficulties of unprovoked testing. In a theoretical example calculation, it assumes that of 100,000 people tested, 50 are actually infected. A test that does not recognize two percent of all infections as such (false negative rate) would work correctly in 49 out of 50 infected people.

In addition, Lühmann assumes in their calculation that the test will work for every hundredth non-infected person. According to the Robert Koch Institute (RKI), the actual value for the false positive rate of the PCR tests cannot be given for Germany as a whole. In the case of the 99,950 uninfected people in Lühmann's calculation example, the tests would therefore show a correct result in 98,951 cases. 999 times, however, the test would be (false) positive. That means: a total of 1048 people tested would get a positive result. Of these, however, only 49 would actually be corona infected, i.e. just under five percent.

If you were to test on an occasion-specific basis, i.e. if there is a justified suspicion of corona, the proportion of correct positive tests would be significantly higher, so the argumentation. It must be clear: The basic assumptions are purely theoretical and are not based on actual information, for example on the error rate when dealing with the test systems currently in use - simply because such data are not publicly available.

Could a noticeable proportion of the reported infected people actually not be infected at all?

No, says Christian Drosten, the head of the national consulting laboratory for coronaviruses. "The result of a laboratory test is always a diagnosis, never a raw test result," he explains when asked by the German Press Agency. And that makes a big difference. “Particularly in the case of positive test results, an additional test is always confirmed (additional gene site). In this way, the occurrence of false positive diagnoses is practically prevented to zero, ”explained the virologist, to whom the Association of Accredited Laboratories in Medicine also refers to this question.

The statistics of the RKI also contain diagnoses, not raw test results, said Drosten. Claims about the unreliability of PCR test results have almost always been based on confusing technical results with medical findings.

Do laboratories always test twice if the results are positive?

The German Press Agency asked several large laboratories as an example. The specific answer was given by Synlab, a provider that, according to its own information, currently carries out up to 80,000 tests per week. Synlab writes that it does not test for multiple gene sites by default. Also, not every positive test result is confirmed with an additional test. In view of the expertise and quality of the tests, this is no longer necessary.

The laboratory operator Bioscientia explains on its website that the tests are looking for three virus genes. Therefore, the so-called total specificity adds up to 99.99 percent. According to this, one out of 10,000 non-infected people gets a false positive result, wrongly believing that he is infected.

Why is it so important that there are as few false results as possible?

For decisions about stricter measures to combat the pandemic or to relax, politicians also look to the numbers of reported infected people. Wrong data could therefore have direct consequences for the practical everyday life of every individual.

Incidentally, false negative results - i.e. unrecognized infections - can also have consequences: From an epidemiological point of view, they are even much more dangerous because potential super-spreaders can continue to move freely and thus infect many more people.

Conclusion

Current sample calculations for false positive evidence are based on purely theoretical basic assumptions. The real values ​​- for example for the error rate when carrying out the tests - can hardly be estimated with the data available to date. According to Christian Drosten, it can be ruled out that, as a statement circulates on the Internet, the majority of the infections currently recorded do not actually exist. “This risk is numerically irrelevant,” he emphasizes. However, Dagmar Lühmann points out, one requirement is very justified: It is essential that valid study data on test quality are scientifically published.