Key challenges faced by Covid-19 tests | Opinion
Both contestants in the US Presidential race have been tested for Covid-19 very recently. One has been hospitalised after a positive test result, despite his being protected by regular testing for himself and anyone who he meets at home, work or a rally.Updated: Oct 05, 2020, 06:57 IST
Both contestants in the US Presidential race have been tested for Covid-19 very recently. One has been hospitalised after a positive test result, despite his being protected by regular testing for himself and anyone who he meets at home, work or a rally. The other has been cautioned by experts that a negative viral test result in his case is no assurance that he is free of the virus and that he should be tested again. In the same week, the latest antibody survey in Delhi was reported to show a decline in the percentage of persons showing evidence of past infection with Covid-19, compared to the previous month! How do we explain these oddities related to the tests?
Virus detection tests for the Saars-CoV2 virus are principally of two types. The molecular tests depend on nucleic acid amplification from the virus which is usually extracted from the nose or throat through a swab. The test is then performed in a qualified laboratory using the Real Time RT-PCR (Reverse Transcription-Polymerase Chain Reaction) or an equivalent process. This takes several hours and reports are received after a day or more- sometimes with even greater delays, as reported in the US. Frustration at these delays has led to the emergence and use of the Rapid Antigen Tests (RAT). These detect known antigenic fragments of the virus and can yield reports within half-an-hour. Easier to perform and earlier to report.
However, neither test is perfect for detecting the virus in all the persons who are infected. RT-PCR can miss between 30-40% of cases, mainly because the sampling with a swab was done too early in the infection when the virus replication is not prolific or too late in the infection when the viral counts are dwindling. Also, the swab collection technique may have been defective. These can be problems with the RAT too. Compared to the RT-PCR, the rapid test misses even more cases. While these estimates vary, between the manufacturers who test in laboratory conditions (technical sensitivity) and those who test in real world conditions (clinical sensitivity), there is a danger of missing half the infected cases with RAT. Hence the recommendation for repeated testing if the report is negative. It is now known that the US President and the persons he met were screened only by RAT.
Antibody tests are employed for random surveys of the population through sampling survey techniques, to determine what fraction of the population has already been exposed to the virus and has recovered with a detectable level of immunity in the blood. The test does not tell us if these are neutralising antibodies capable of arresting the virus nor does it give an indication of T cell immunity which is the other, and probably longer lasting, component of the immune response.
Unfortunately, the antibody response to the Covid-19 virus appears to fade away in three months. Hence, persons infected during March to May may not be identified by a survey in September. The sampling design, and the participant responses, too can produce variations in the nature of persons tested. Also, every estimate (point estimate) needs to be reported with 95% confidence intervals which represent the range of probabilities which that point estimate represents in a given sample size. Unless the sample is large, the confidence intervals can be wide and overlap even for seemingly different point estimates.
While the concern with virus detection tests is about false negative results, the challenge with antibody surveys lies in estimating the false positive results.
Other corona viruses can present cross-reactive antibodies and recently some overlap with Dengue too has been reported. When the lab test is used for a large population survey, where prevalence of the infection is lower than the clinical setting where it has been validated, the number of false positive results get amplified.
The estimated prevalence will be higher than the truly infected proportion of the population. This error band can vary from survey to survey, depending on the actual prevalence and the sample size. Population surveys can still provide a broad understanding of how the epidemic has advanced over time, if we don’t regard each point estimate as the absolute truth. One thing is for sure- don’t take anything for granted in these Covid times!
(The author, a cardiologist and epidemiologist, is President, PHFI. He is the author Make Health in India: Reaching a Billion Plus. Views are personal.)