Related articles in Covid-19

Covid-19 Test Accuracy

How accurate are the ‘lateral flow’ or ‘rapid flow’ Covid-19 test kits used by schools and workplaces and now being offered to people across the UK? BuDS’ expert team has looked at a range of scientific evidence so that we can give you some reliable, fact-checked, information. This information relates to England only.

This is a long article, and deals with complex material and statistics. If you would prefer to read a shorter and simpler version of this article, please use this link: https://buds.org.uk/covid-19-test-accuracy/ or use the contents below to jump to our ‘What Does This Mean For Me’ summary.


 

THREE TYPES OF TEST

 There are three types of tests for Covid-19. All the tests use the same basic sample. This sample is taken from people by putting a swab into your mouth and nose to collect saliva and mucus. Because Covid-19 is a virus that you breathe in, it is most likely to be found in your mouth and nose.

The first type of test is the RT-PCR (reverse transcriptase polymerase chain reaction) test. This is a test using an advanced machine which analyses and catalogues the genetic material in the sample, then sees whether any matches the Covid-19 virus genetic material. The test is done only in laboratories by trained scientists. This means that samples have to be sent to the lab in the post. The RT-PCR test also detects which variant of the Covid-19 virus is present in the sample plus a lot of other useful material about the virus. These tests are most often found as the postal test kits that can be requested from the Government website.

The second type of test is a LAMP (loop-mediated isothermal amplification) test kit. This is not a laboratory test, but a kit rather like a home pregnancy test designed for use in community settings like schools, prisons, workplaces, etc. LAMP test kits also work by looking for genetic material belonging to Covid-19. A chemical in the kit changes colour when genetic material from Covid-19 is discovered, showing that there is Covid-19 in the sample. LAMP test kits are not much used yet as they were only approved for use earlier this year. They are made by companies called OptiGene and Oxford Nanopore.

The third type of test is a lateral flow test kit or LFT test kit. Again, this is a kit rather like a home pregnancy test – but designed for use in community settings. The kit has a strip coated with a protein-based chemical (built around antibodies which react to Covid-19) which reacts and changes colour when it has contact with a protein found within the Covid-19 virus. LFT tests have been used in England for some time now. The main LFT test kit used in England is made by a company called Innova but there are also kits made by Abbott Panbio and SureScreen Diagnostics.


 

SENSITIVITY, SPECIFICITY, FALSE NEGATIVES AND FALSE POSITIVES

Covid-19 tests show whether you are infected with Covid-19 or not. No test is 100% accurate, so scientists look at how accurate they are. They look at a number of things:

  1. If the test shows you ARE infected with Covid-19, how likely is the test to be right? This is known as the test’s ‘sensitivity’.
  2. If the test shows you are NOT infected with Covid-19, how likely is the test to be right? This is known as the test’s ‘specificity’.
  3. Does the test find the virus in samples from people who don’t have the virus? Finding that Covid-19 is present in samples from people who were not infected is called a ‘false positive’ result. This means the test is falsely showing that the sample is positive for Covid-19.
  4. Does the test not find the virus in samples from people who do have the virus? Finding that Covid-19 is not present in samples from people who were infected is called a ‘false negative’ result. This means the test is falsely showing that the sample is negative for Covid-19.

 

IT’S ALL ABOUT THE PEOPLE, NOT THE TEST

It’s important to remember that there is a difference between how a test performs in ideal circumstances and how a test performs in the real world. Ideal circumstances might be when the test is used in a lab by fully trained staff using samples taken by trained professionals, while in the real world the test kit might be used in a school by a quickly-trained school secretary using a sample taken by a teenager. The quality of the sample and how well the test is done is much more important than the scientific accuracy of the test when looking at overall accuracy.


 

RT-PCR TESTS

There is a lot of conflicting evidence about the accuracy of RT-PCR tests, and this has been exploited by people wanting to spread fake news or deny the reality of Covid-19. BuDS has sifted through the evidence to try to give you a balanced view.

Here are some facts about the RT-PCR test about POSITIVE results:

  1. The PCR test machine itself is very accurate indeed. If samples are prepared by highly trained technicians, and the machine says that the sample is positive for Covid, the machine will be right in about 99% of cases.1
  2. If you send a high-quality sample to a lab for RT-PCR test, and the machine says that the sample is positive for Covid, the machine will be right in about 95% of cases.2 A high-quality sample might be one taken by a health care professional or by someone at home strictly following the instructions.
  3. If you look at RT-PCR tests in general, they correctly detect the virus in about 70% of cases.3 This is not because the test machine is inaccurate, but because the samples fed into the machine are of lower quality, because they have been taken by unskilled people.

So, if you send a sample off in the post for an RT-PCR test, and the test comes back positive, you can be 95% sure that you have been infected with Covid IF you have done the mouth and nose swab carefully. If you have been less careful with your mouth and nose swab, there is still a high chance that you are infected with Covid and you should definitely self-isolate (along with your household) so that you do not give Covid to other people. But, if you haven’t been careful with your sample, you may like to do another RT-PCR test and this time take more care or get a sample taken at a healthcare setting by a professional.

Here are some facts about the RT-PCR test about NEGATIVE results:

  1. If you send a high-quality sample to a lab for RT-PCR test, and the machine says that the sample is negative for Covid, the machine will be right in about 76% of cases.4 A high-quality sample might be one taken by a health care professional or by someone at home strictly following the instructions. If a low-quality sample is sent to a lab, the probability of the machine being right falls
  2. It is possible for a test to falsely show that a sample is negative for Covid-19 (i.e. the person who provided the test actually had Covid-19 when the test says they didn’t). This is known as a ‘false negative’ result. If a high-quality sample is provided, around 2% of results may come back as ‘false negative’.5 If a low-quality sample is provided, up to 29% of results could come back as ‘false negative’.6 This variation is only based on the quality of the sample, and not the test machine itself.
  3. Just because a result has come back as negative, it does not mean that you shouldn’t isolate. Even if the machine was entirely accurate and correct, there is still a chance a person could be infected with Covid-19. This is because the accuracy of the test is dependent on a high concentration of Covid-19 antibodies, which are often not present in the first 3-7 days of infection.7 Hence if you take a test in this time period and it comes back negative, but you still have symptoms of Covid-19, you should continue to isolate and take another test to confirm the diagnosis.

So, if you send a sample off in the post for an RT-PCR test, and the test comes back negative, you can be 76% sure that you have not been infected with Covid IF you have done the mouth and nose swab carefully. If you have been less careful with your mouth and nose swab, there is still a high chance that you are infected with Covid if you have symptoms – you should definitely self-isolate (along with your household) so that you do not give Covid to other people. If you do not have symptoms, there is still a chance that you are infected, but this is less likely. If you haven’t been careful with your sample – whether you have symptoms or not – you may like to do another RT-PCR test and this time take more care or get a sample taken at a healthcare setting by a professional.


 

 LAMP TESTS

The LAMP (loop-mediated isothermal amplification) tests have been in development for the last few months, and only recently were approved for some community use. As such, there is not much real-world information on how accurate they are. However, based on small-scale lab testing, they appear to be “at least equivalent to RT-PCR” tests8. They also appear to be potentially more versatile in community use, due to a reduced complexity in the processing stages.9,10 However, more testing is needed in both labs and the real world for this to be proven fully, and as such BuDS will not be going into more detail about these tests at present.


 

LATERAL FLOW TESTS (LFTs)

As with RT-PCR tests, there is a lot of conflicting evidence about the accuracy of lateral flow tests (LFT). In brief, the Government ordered Lateral Flow test kits from an American company, Innova, before full analysis of these tests had been completed. It then completed its own testing at the Government laboratory at Porton Down in collaboration with Oxford University, plus in a separate supervised study run by Oxford University. Both these tests were run under the supervision and direction of the Department of Health and Social Care. However, the results obtained in these tests differ widely from the community results obtained in real-world use of the tests, making the true accuracy of LFTs hard to gauge. BuDS has sorted through the evidence, and will try to give you the most balanced view of the situation based on all the scientific data we have.


 

HOW ACCURATE DO LFT’S NEED TO BE?

First, it is important to know how accurate lateral flow tests are meant to be, according to the World Health Organisation (WHO). WHO guidance requires lateral flow tests to have a minimum sensitivity of 80%, and a minimum specificity of 97%.11,12 This means that if you are infected with Covid-19, at least 80% of tests must show this if you use them. It also means that if you are not infected with Covid-19, the test must correctly show this in at least 97% of cases.


 

THE DATA

There are 2 main sources of information relating to the accuracy of lateral flow tests. One of these is data from the Government’s own testing of lateral flow tests they had already ordered, while the other is data from organisations looking at test accuracy in the real world.  BuDS has looked at studies from both source pools, and will summarise the data below.

The Government has carried out 2 major studies into the accuracy of lateral flow tests (as of March 2021). One of these studies was carried out at PHE Porton Down, a military laboratory usually associated with chemical testing and viral research.13 The other study was carried out by Oxford University under the guidance of the Department for Health and Social Care.[14 These two studies were carried out mainly under lab conditions, and assumed ideal conditions were in place for viral detection when evaluating the test examined. Both sets of results relate only to the Innova test. The results of these studies are summarised below.

  1. Sensitivity: the Porton Down study recorded an average sensitivity in their tests of 79.2% under lab conditions, 73% in trained health professionals, and 57.5% accuracy in real-world community testing.15 The Oxford University study recorded an average sensitivity of 50.1% overall.16 Notably, both sets of results are at least 30% below the required sensitivity set out by the WHO.17
  2. Specificity: the Porton Down study recorded an average specificity in their tests of 99.68% under lab conditions, and 99.61% in the community.17 The Oxford University study recorded an average specificity of 99.72% under lab conditions, and 99.7% in the community.19However, this assumed that the accuracy of the RT-PCR test used to verify the lateral flow tests was 100%.20 As a result, the average specificity in this study could be as low as 76% (but this was not verified in the official study). Both of these studies returned results that were higher than the WHO requirement, assuming that the RT-PCR accuracy did not affect the Oxford study results.
  3. False positives: the Porton Down study had 0.32% of their tests come back as false positives under lab conditions, and 0.39% in the community.21 The Oxford University study recorded no false positives in either circumstance.22
  4. Identification and failure: the Porton Down tests saw an average of 5.4% of tests fail to work properly (between 0.65% and 16.8% depending on circumstance).23 The tests examined had a 95.5% identification rate under ideal conditions, but this fell to 60% when non-ideal conditions were used.24 The Oxford University tests did not report on the identification or failure rates of the tests examined.25

Overall, the results of the Government’s studies of lateral flow tests show that while the tests are very good at detecting when a person does not have Covid-19, they are significantly below expectations when detecting when a person has Covid-19. In addition, an identification rate of 60% under non-ideal conditions (i.e. when there is not a lot of Covid-19 antibodies present on the sample being tested) means that only 60% of the cases were actually detected, of which only 50%-70% were correctly analysed.

The other main data source we have for the accuracy of lateral flow tests is from large-scale testing carried out by two independent organisations. BuDS has looked at 2 major studies, one from a group called the Cochrane Library in October to November 2020,26 and the other from the University of Birmingham in December 2020.27 The University of Birmingham study is particularly interesting, as it is the first entirely real-world test into test accuracy. The test was carried out in the student population, in the same way that thousands of people across the country are now being tested for work and study, making it a very good representation of real-world accuracy. Both studies looked at the Innova test. The results are again summarised below.

  1. Sensitivity: the Cochrane Library study recorded an average sensitivity of 72%, (between 34% and 88%), with the Innova tests recording an average of 58%.28 The University of Birmingham study recorded an average sensitivity of 3.2% – a shocking figure.29 All these figures are again notably below the WHO requirement of 80%.30
  2. Specificity: the Cochrane Library study recorded an average specificity of 99.2%, which fell to 98.9% in people with no symptoms of Covid-19.31 The University of Birmingham study recorded an average specificity of 99.74%.32 Both these values are above the value of 97% required by the WHO.33

Overall, the results from the independent studies once again show that lateral flow tests are very good at detecting when a person does not have Covid-19, but they are significantly below expectations at detecting when a person has Covid-19 – to a far greater extent than the Government studies suggested. This may well be because the independent testing was carried out in more realistic conditions than the closely controlled labs the Government used, which will inevitably affect the results taken.


WHAT DOES THIS MEAN FOR ME?

The results of the studies above clearly show that while lateral flow tests are good at detecting when a person does not have Covid-19, they are terrible at detecting when a person does have Covid-19. The accuracy of the tests falls significantly in samples with a low viral load (i.e. when only a small amount of Covid-19 antibodies are present in the sample), which is the situation in the early and late stages of infection as well as in poorly-taken samples. The recommendation of both independent studies was that if lateral flow tests are to be used, they should be used very regularly (much more often than twice a week, as the Government are currently promoting), and that a negative result should never be used to guarantee that someone is free of infection.34,35 However, this is exactly how negative results are being used.

If you receive a positive test result, it is very likely that you are infected with Covid-19. You should isolate immediately, and take a RT-PCR test to verify the result.

If you receive a negative test result, it is possible that you are still infected with Covid-19 due to the sensitivity of the lateral flow tests missing edge cases. You should continue to be very careful, keep your distance from others and wear a mask wherever possible. If you have any symptoms of Covid-19, or are at high risk, you should also take a RT-PCR test to verify the negative result. You should also take a RT-PCR test to verify the result if you live with someone who is high risk or work in high-risk settings, such as in education or health.


 

MORE INFORMATION

To see all BuDS’ articles about Covid-19, use this link: https://buds.org.uk/category/iag-covid-19/

To read a shorter summary of this article, use this link: https://buds.org.uk/summary-of-covid-19-test-accuracy/

To learn more about Long Covid and its impact, use this link:  https://buds.org.uk/long-covid-july-2021-update/

To learn more about the people who are more likely to die or have serious illness if they catch Covid-19, use this link: 

FINALLY

Please share this article on social media, but always credit BuDS. If you need help or support or you’re anxious about Covid-19, BuDS is here for you. Please e-mail buds-support@buds.org.uk, call 01494 211179 (voicemail) or message us and we’ll do all we can to help.


 

REFERENCES

[1] https://www.bmj.com/content/bmj/369/bmj.m1808.full.pdf

[2] Ibid.

[3] Ibid.

[4] Ibid.

[5] Ibid.

[6] Ibid.

[7] Ibid.

[8] https://pubmed.ncbi.nlm.nih.gov/33848785/

[9] Ibid.

[10] https://doi.org/10.3389/fmed.2021.627679

[11] https://www.who.int/publications/i/item/9789240017740

[12] https://www.bmj.com/content/372/bmj.n823

[13] https://www.ox.ac.uk/sites/files/oxford/media_wysiwyg/UK%20evaluation_PHE%20Porton%20Down%20%20University%20of%20Oxford_final.pdf

[14] https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/968095/lateral-flow-device-specificity-in-phase-4.pdf

[15] https://www.ox.ac.uk/sites/files/oxford/media_wysiwyg/UK%20evaluation_PHE%20Porton%20Down%20%20University%20of%20Oxford_final.pdf

[16] https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/968095/lateral-flow-device-specificity-in-phase-4.pdf

[17] https://www.who.int/publications/i/item/9789240017740

[18] https://www.ox.ac.uk/sites/files/oxford/media_wysiwyg/UK%20evaluation_PHE%20Porton%20Down%20%20University%20of%20Oxford_final.pdf

[19] https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/968095/lateral-flow-device-specificity-in-phase-4.pdf

[20] Ibid.

[21] https://www.ox.ac.uk/sites/files/oxford/media_wysiwyg/UK%20evaluation_PHE%20Porton%20Down%20%20University%20of%20Oxford_final.pdf

[22] https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/968095/lateral-flow-device-specificity-in-phase-4.pdf

[23] https://www.ox.ac.uk/sites/files/oxford/media_wysiwyg/UK%20evaluation_PHE%20Porton%20Down%20%20University%20of%20Oxford_final.pdf

[24] Ibid.

[25] https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/968095/lateral-flow-device-specificity-in-phase-4.pdf

[26] https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD013705.pub2/full

[27] https://www.medrxiv.org/content/10.1101/2020.12.01.20237784v2

[28] https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD013705.pub2/full

[29] https://www.medrxiv.org/content/10.1101/2020.12.01.20237784v2

[30] https://www.who.int/publications/i/item/9789240017740

[31] https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD013705.pub2/full

[32] https://www.medrxiv.org/content/10.1101/2020.12.01.20237784v2

[33] https://www.who.int/publications/i/item/9789240017740

[34] https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD013705.pub2/full

[35] https://www.medrxiv.org/content/10.1101/2020.12.01.20237784v2