Yves here. Ignacio describes some of the promising methods for Covid-19 diagnosis, and the considerations involved in scaling up their use. He explains why different countries may wind up opting for different tests, or different sequencing and frequency of tests. Sadly, I would assume those variations will complicate, and perhaps impede, aggregating findings across nations.
By Ignacio Moreno Echanove
I will try to focus on Covid-19 diagnosis not as a clinical tool but as an epidemiological surveillance tool badly needed If we are one day to return to whatever can be the new normal. This means that besides being accurate, the methodology used must be suitable for extensive geographical deployment (particularly sampling), early detection (even before symptoms appear), scalable (automation), and avoid delays in the flow of information.
Regarding accuracy the requisites are the same as for clinical diagnosis: The method must be precise (reproducible), sensitive (positive/infected > 85%) and specific (negative/non-infected closest to 100%). All those measures of diagnostic efficacy are relative, obtained by comparison to what is considered the gold standard for a particular pathogen. The gold standard for Covid-19 detection is based on the amplification of viral nucleic acids by RT-PCR tests (NAATs). In the case of NAATs for virus diseases, sensitivity can be measured as the number of viral genome copies per test required for 100% positive results. Real Time quantitative RT-PCR methods are quite sensitive detecting as few as about 3 genome copies in a single reaction (1). Well-designed NAATs kits are also highly specific and spurious amplification (reaction artifacts) or amplification of genomes other than SARS CoV 2 can be ruled out in best lab-conditions (1).
Being true that NAATs are quite sensitive and specific, and can be automated, there are problems with the “A” for amplification. Being so sensitive, you can amplify virus genomes obtained by sampling infected individuals and/or present in the reaction by environmental contamination anywhere in the process. This results in false negatives, lower specificity. Thus, precautions are needed in the whole testing process, particularly during the sampling because lab conditions are better controlled than sampling sets and internal controls can detect or rule out lab contamination.
Besides, such high sensitivity doesn’t necessarily translate into an optimal value for sensitivity in the real world (positives/infected). Problems with sampling or transport/storage, as well as the highly variable extent/state of the infection in each subject tested may reduce sensitivity very much. Sampling is probably one of the most important factors reducing the sensitivity of the method and should be done by trained people. Sampling is usually done by swabbing the nasopharyngeal and/or the oropharyngeal tracts (Upper Resp. Tract, URT) because it is convenient and fast. Some studies indicate that the combination of both (nasal+oral) gives best results (2). However, Covid-19 infections in the URT may not be very extensive and constant in time because there have been several reports indicating, for instance, that symptomatic individuals showed negative only to test positive in a second test.
Also, it looks like there is, in the population at large, wide variability in URT infection development that will be matched by similar variability in sensitivity. Information on infection dynamics in the URT is scant probably because this is not that important from the clinical point of view. Notwithstanding, this is critical from the epidemiological point of view and the ability of SARS CoV 2 to infect and replicate in the URT is what distinguishes it from the not so infectious SARS1.0 (3). This paper shows that the window for accurate Covid-19 detection in URT swabs in symptomatic subjects by NAATs is between 6-14 days after symptom appearance, with best results between 7-11 d.a.s.a. This is a narrow opportunity for test accuracy when you are testing many individuals.
Very little is known about symptomless or nearly symptomless individuals but it has been shown that they can prove positive in tests and be readily infected (3). All this translates into lower sensitivity than expected and some puzzlement amongst physicians. Besides, there are mistakes: Labelling and sample tracing mistakes aren’t that rare particularly in stressing situations. To complicate the picture symptoms reported during the first week are quite variable but anyone feeling fever, cough, cephalgia, rhinitis or sinusitis should be tested. Widespread allergies compound to make symptoms more unreliable.
Even with the methodological limitations exposed, NAATs that have been well designed, validated, and performed by properly trained (and protected) personnel are quite a useful tool for epidemiological surveillance as South Korea has demonstrated. To get an idea on what is needed, how the sampling is done, and how the HC system must be protected I recommend watching this video: https://www.youtube.com/watch?v=b0D8b1WZJTM, particularly go to minute 14 to see correct sampling procedures in a drive-through testing set. You will see the nasopharyngeal swab is much nastier than that from the oropharynx.
And an important conclusion one can draw from this video is that this is not all about the diagnostic method. Effective surveillance and containment require additional measures such as social distancing, greased reporting, fever surveillance and the population at large must collaborate in the surveillance and containment. Remarks by Prof. Schaffner at about min 3:30 of the video are quite correct. My conclusion is that the limitations of the method can be overcome with numerous, frequent and repeated tests that span well beyond reported symptomatic cases. Also, isolation of positives preventing home contagions might be necessary or at least desirable in those cases where self-isolation is difficult. In any case, the known difficulties with NAATs has sparked research on alternative methods.
I will skip immunological assays (detection of antigens or antibodies) because more time is needed for kit production and validation and many of these have drawbacks for deployment in epidemiological surveillance (5). Some are not sensitive enough while others are sensitive but not easily scalable etc. Detecting antibodies is important from a clinical point of view and also from the epidemiological point of view to have an idea on the state of immunity throughout the population but there are issues to discuss that would make this too long.
Yet, there are some other methods based on NAT. There are NATs that avoid amplification and are highly sensitive, for instance CRISPR based methods (a couple mentioned in 5). Other set of methods are based on RNA sequencing. Next generation sequencing, HiSeq, Single-Cell RNA sequencing, and/or single-cell RNA sequencing are rapidly evolving methods that have been developed to study and diagnose an array of non-infectious and infectious diseases. These are based on the sequencing of all cellular mRNAs, the “transcriptome”, that will include viral RNAs if infected. The sequences obtained are in silico filtered to remove human sequences and then you may obtain specific viral RNA sequences. You can run a ‘blast’ (tool for massive comparison of sequences with existing databases) with the sequences obtained and get unbiased results: this is certainly Covid-19, or no, this is flu H5N1.
These kinds of tests, though complex and requiring further training, can be easily automated, scaled up (“multiplexable” in the jargon) and some of them do not depend on the availability of processing kits. Investment in costly equipment is needed but it is already available in some research centres.
These methods, compared with real-time qPCR offer several advantages: the pathogen is simultaneously detected and genotyped, internal controls ensure unambiguity better that with qPCR and there isn’t amplification associated risks. I cannot show links but according to personal communications the sensitivity can be higher by an order of magnitude in terms of number of infected cells required for virus detection so these methods may allow for earlier contagion detection compared with NAATs. Regarding speed, results can be obtained in a couple of hours, similar to NAATs in this sense.
The problems are that standardization of the procedures is needed, followed by validation against the gold standard, and in many regions of the world there is not equipment neither trained personnel available. There are already some commercial systems based on these techniques using slightly different platforms only for research purposes. To my knowledge, some are in the validation process for Covid-19 in at least four countries. This epidemic could see the rise of these techniques as the new gold standard for clinical diagnostic of virus diseases.
Anyway, the South Korean experience, as well as some other and more negative experiences –for instance Spain is running out of imported processing kits for NAATS– suggests that each country will make decisions to ensure they have the necessary resources to keep their surveillance programs going on without depending on technologic platforms developed by possibly unreliable providers. This suggests that we may end with a mosaic of diagnostics methods developed and validated in each region.
- Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-PCR
- Which sampling method for the upper respiratory tract specimen should be taken to diagnose patient with COVID-19?https://www.ncbi.nlm.nih.gov/pubmed/32166939 (abstract in English)
- Virological assessment of hospitalized patients with COVID-2019. https://www.nature.com/articles/s41586-020-2196-x_reference.pdf
- Alert for non-respiratory symptoms of Coronavirus Disease 2019 (COVID-19) patients in epidemic period: A case report of familial cluster with three asymptomatic COVID-19 patients. https://onlinelibrary.wiley.com/doi/epdf/10.1002/jmv.25776
- Coronavirus tests: researchers chase new diagnostics to fight the pandemic. https://www.nature.com/articles/d41586-020-00827-6