Tests to determine if someone has a current infection are used in several contexts. They are used to diagnose or screen for infections to allow decisions to be made about clinical treatment and subsequent actions such as whether someone or their contacts need to isolate. They are also used as a research tool so that scientists and public health bodies can monitor the prevalence and spread of infections in the population, defined regions, communities or specific groups. Examples of these uses include:
- Confirming a clinical diagnosis. Someone admitted to hospital with suspected COVID-19 is tested to confirm whether they have the infection – this then guides subsequent decisions about where and how a patient will be treated.
- Establishing if someone might be infectious. Testing allows people with the infection (and who may not have symptoms) to be identified and isolated to reduce the spread of the virus. For example, regular testing takes place in high-risk settings, such as among the staff and residents of care homes.
- Research. The Office for National Statistics’ COVID-19 infection survey reports national data on the current rates of infection in the population using tests randomly carried out on a large sample of people.
Tests to detect current infections
There are two main types of tests that can detect the presence of the SARS-CoV-2 virus:
- Molecular tests detect viral genetic material called RNA.
- Antigen tests detect proteins found on the surface of the virus.
Tests work by detecting the presence of either of these elements in a sample from a person. This is usually from fluid collected on swabs taken from places in the respiratory system where the virus is likely to be found (the virus can also be detected in stool and in blood). These samples are typically taken from the upper parts of the respiratory tract, commonly the nose and throat, but in healthcare settings can be taken from locations deeper down. There has also been recent interest in developing tests that can analyse saliva samples. Nose and throat swabbing can be uncomfortable, so approaches to develop tests that are less invasive are of interest, especially for children, in contexts where people will be collecting samples themselves or where the frequency of testing is high.
The type of sample taken and when it is collected is important because the level of virus present varies in different parts of the body, changes over time, and may differ with the severity of the infection and the person’s age. Virus can be detected in respiratory samples from the onset of symptoms for up to 2 weeks. Virus can also be detected in infected people who have no symptoms at all, or later in the course of an infection, by which time they are unlikely to be infectious any longer.
Test results can be processed from samples in minutes to hours. This depends on the type of test used and the capacity of the wider testing infrastructure at any given time. At present, all national testing programmes use tests that require sending the samples to laboratories, where trained staff will process and interpret the results. There is significant research activity and government interest and investment in developing test technologies so that they can:
- Be processed closer to the location of the person being tested (such as in a care home).
- Give faster results.
- Use less invasive sampling techniques.
- Be produced in larger volumes and at lower cost so that they can be used on a mass scale.
Currently, tests with these features are only being used in research projects in the UK. These are discussed later.
Molecular tests
Different tests use one of several techniques to identify if a sample contains SARS-CoV-2 genetic material.
- RT-PCR: Reverse Transcription Polymerase Chain Reaction is a widely used laboratory method. It uses a technique and special equipment to increase the amount of genetic material from the sample so that it can be detected. These tests tend to have high sensitivity, so they make good diagnostic tests, and therefore have been the mainstay of COVID-19 testing in the UK. Test samples are sent to and processed in NHS Trust laboratories, national public health agency laboratories and the UK Lighthouse Labs Network (a network of diagnostic centres focused on COVID-19 testing).
Research is ongoing to develop tests that can detect viral genetic material without using RT-PCR, with more portable equipment and on other sample types, such as saliva. Newer technologies may allow for the processing of samples more quickly and without the need for laboratory processing. In theory these technologies could deliver results within an hour, with equipment that can be sited locally to the where the person is being tested, for example in a school or care home. Currently these technologies still require special equipment and a trained operator to process and interpret the results.
- RT-LAMP: Reverse Transcription Loop-mediated isothermal AMPlification is a technique that is being adapted to work without the need for laboratory processing and is the focus of development of some new tests. Some test manufacturers report that test results can be processed within 15 minutes.
- CRISPR: another approach that uses enzymes to detect the presence of virus, with some labs reporting that their tests can deliver results within about 40 minutes.
Antigen tests
These tests detect if a sample contains proteins that can identify the SARS-CoV-2 virus. The test equipment or kit contains antibodies that bind to the viral protein if it is present in the sample. A positive result can then be visualised by seeing a fluorescent glow or a dark band on the test kit. These tests do not necessarily have to be carried out in a laboratory. This type of test can give fast results without the need for any laboratory processing or analysis, and is similar to how a pregnancy test works. Antigen tests can be made cheaply and so are well-suited to being used in very large quantities. The company Abbott has developed an antigen test that was approved in August by the US Food and Drug Administration. It is worth noting that the FDA makes it clear that a negative result does not necessarily rule out infection. Data from Abbott reports that sensitivity and specificity is 97.1% and 98.5%, respectively (see How reliable are tests? for definitions). As is expected it has lower sensitivity than PCR, so while it is not a good test to inform decisions about caring for a patient in a hospital, this and antigen tests like it could be useful in population surveillance.
So far antigen tests for COVID-19 have been designed to only be used by a trained operator, who will take the sample, process it if needed, and interpret the result. In general antigen tests tend to be less sensitive than molecular tests, so if someone has a small amount of virus in their body the test might not pick it up. This could lead to a false negative result (someone has an infection, but the test says that they don’t). Researchers are working to improve the sensitivity of these tests. However, some researchers suggest that test sensitivity should come second to the ability to test frequently and obtain results quickly, in the context of infection surveillance.
Using tests to find out if someone is infectious
The amount of virus in the body (viral load) changes over the course of an infection. The viral load in the respiratory tract peaks in the first week and can be detected for up to 2 weeks. Research indicates that live virus (which can cause an infection) is unlikely to be present 10 days after symptoms begin. However, viral material that cannot cause an infection can still be detected over an even longer period, sometimes up to 2–3 months. Therefore, highly sensitive tests such as RT-PCR can potentially detect viral genetic material from someone after the point when they cease to be infectious. In this case less sensitive antigen tests may be a better option for screening approaches, since they will detect people with higher levels of virus who are more likely to be infectious. This is not straightforward because there is uncertainty about what viral load constitutes infectiousness.
Some tests can offer more information than simply indicating if the virus is present or absent. For example, they may provide information about the quantity of virus that is present.
How reliable are tests?
As with any diagnostic test, data about how confident we can be about their accuracy and reliability is crucial. Depending on the context in which the test is used, different test characteristics may be more important than others. The accuracy of testing also depends on what proportion of the population have an infection at any given time. Although high quality tests are available, none can claim 100% accuracy. This is because there is no gold standard reference to diagnose COVID-19 and no agreed shared standard against which manufacturers can report the comparative performance of their tests. This has meant that public health agencies have had to develop their own reference standard, and then evaluate the performance of commercial tests against it, in order to give governments a clear idea of how tests perform.
Sensitivity and specificity of tests
The ability of a test to detect very small amounts of virus is important and this will vary between tests. Some samples may contain less viral material than others and the amount of virus in the body changes as an infection progresses. Tests also need to be able to react only to the SARS-CoV-2 virus and not to other viruses that may also be present in a sample, especially other coronaviruses.
When the accuracy of tests is discussed the most important terms used are:
- Sensitivity: the proportion of people with SARS-CoV-2 infection who test positive. A test with sensitivity of 95% would mean that 5 in 100 people who have COVID-19 would test negative (false negative). They have an infection, but the test says that they don’t.
- Specificity: the proportion of people without SARS-CoV-2 infection who test negative. A specificity of 90% means that 10 in 100 people who are not infected still test positive (false positive). They do not have an infection, but the test says that they do.
These figures are cited by manufacturers when they describe the accuracy of their tests. However, several factors influence overall accuracy when tests are used operationally in real settings outside the controlled conditions of a laboratory. When levels of infection in the population being tested are high, a test with a high level of sensitivity will be very good at identifying people with an infection but less good at detecting people that do not. Conversely if the level of infection in a population is low, then a higher specificity becomes more important because it will be better at identifying the people that don’t have the infection than it is at detecting the people that do.
Implications of test performance for how they are used
These differing characteristics mean that no one test is best suited to all the possible purposes they can be used for. The way in which they will be used also has implications for the accuracy of results. Tests with high sensitivity work well when there is a high chance that the person is infected. An example of this would be the testing approach used in the test and trace programmes across the UK that seek to confirm a diagnosis of suspected COVID-19 in people that come forward for testing on the basis of having symptoms, or of having had close contact with an infected person. Specificity is much more important if tests are used to screen very large numbers of mostly healthy people. While there may be a greater degree of tolerance for false positive and negative results in a mass screening programme, the implications of incorrect results are significant because, even for a test with high sensitivity and specificity, large numbers of people would get a false positive (and be required to self-isolate) or a false negative (and potentially go on to infect other people).