Skip to main content

Table 3 Biases and issues in interpretation

From: At what times during infection is SARS-CoV-2 detectable and no longer detectable using RT-PCR-based tests? A systematic review of individual participant data

Domain Details of bias and applicability issues Impact on interpretation of study data
Participants (source of bias) In these studies, the reference test usually incorporates RT-PCR (index test).
• RT-PCR testing is usually a key component of identifying people with SARS-CoV-2 infection.
• Participants will not be detected or included in these studies when SARS-CoV-2 is not present at easily sampled sites and at the time that participants were available for testing.
Unclear how many and what severity of participants with SARS-CoV-2 are not included in studies.
People who do not have a positive RT-PCR test at some point are excluded. This could lead to overestimation of positivity.
Rates of positivity will be inflated as only people with virus accessible for sampling for RT-PCR tests will be included in studies.
Participants (source of bias) Most participants are identified or present based on respiratory tract symptoms such as cough or respiratory distress. Unclear how many and what severity of participants with SARS-CoV-2 are not included in studies.
• Participants will not be detected or included in these studies when less common symptoms or asymptomatic.  
• Participants included will be biased to over-represent people with detectable virus in respiratory tract sampling sites and at times frequently used for testing (post symptom onset or at admission to hospital). Studies will inflate positivity for sampling sites that overlap with sampling sites used in RT-PCR reference testing.
• For example, we identified 30% of participants with RT positivity but with negative results from faecal sampling. However, if participants had only faecal virus, would they have been included in the studies?
Index test: RT-PCR (applicability) • Studies included are likely to use more invasive sampling methods than acceptable in widespread population testing. For example, nasopharyngeal testing is likely in many current studies to be based on long swabs and self test kits. Percentage of people with detectable virus may be overestimated when testing is applied in real-world clinical use and in population testing.
• Studies will use experienced staff to obtain samples, handle, process, and conduct tests.  
• Studies are mostly sampling participants in hospital settings or in specialised research community testing research where sample handling, transport, and storage have been standardised.  
• Variation in RT-PCR kits is minimised as studies are based in few hospitals or limited to a research setting  
Index test: RT-PCR (applicability) Sample RNA extraction methods are designed predominantly for respiratory samples.
• RT-PCR sample preparation kits used are mostly designed for extraction of virus from respiratory samples, not from faecal samples. It is unclear how efficiently these kits extract virus RNA from stool samples.
Percentage of people with virus present in faecal samples and duration of shedding in faecal samples may be underestimated.
Index test: RT-PCR (applicability) RT-PCR tests detect both infectious and inactive (inactive due to immune system or dead) virus. Percentage of people with clinically important detectable virus may be overestimated.
• Few studies link RT-PCR testing to cell assays to test for infectious virus.  
Index test: RT-PCR (applicability) Rate of virus aggregation or clearance by immune system may affect the sampling efficiency at some sampling sites.
• Both the innate and adaptive immune system will aggregate and clear virus particles from bodily fluids. It is not clear what the time scale of clearance or how this influences detection of virus at different sites and at which time points.
Percentage of people with detectable virus may be underestimated.
Index test: RT-PCR (applicability) Reporting of sampling sites and methods is poor.
• Poor reporting may have led to less ideal grouping of sampling in analysis.
• Some studies are likely to use a variety of nasopharyngeal sampling methods depending on the individual participants, but the type of sampling is typically reported at a study level for a particular sampling site.
Percentage of people with detectable virus may be over- or underestimated.
Flow and timing Uncertainty and inconsistencies in time of sampling Percentage of people with detectable virus may be over- or underestimated at particular times.
• Time of symptom onset can be subjective unless based on fever, but some participants do not have fever.
• Time of symptom onset may be different if asked of participants in ICU setting.
• Time of hospitalisation and discharge may be affected by function hospitalisation serves in containment of disease spread. In some studies, the hospitals were also quarantine centres, so participants were hospitalised immediately at onset of mild symptoms rather than restricted to patients needing oxygen.
Flow and timing Clinical cohort within studies changes across time points. Percentage of people with detectable virus may be overestimated at particular later time points as these correspond to participants who were severely ill.
• Participants who have recovered from COVID-19 in most studies are typically not tested after 2 negative tests 24 h apart.
• Many studies only test inpatients at the hospital, so the participants sampled between 0 and 14 days typically have less severe disease than those tested longer
Flow and timing (selective outcome reporting) Some studies only publish IPD data for a selection of people. Available IPD data may not represent a typical spectrum of participants in the different settings (community setting, hospital, ICU, nursing home, prison).
Publication bias Published data is likely to be biased towards publication of research active groups which may not represent typical real world. Percentage of people with detectable virus may be overestimated.
  1. BAL bronchoalveolar lavage, COVID-19 coronavirus disease 2019, ICU intensive care unit, IPD individual participant data, RNA ribonucleic acid, RT-PCR reverse transcription polymerase chain reaction, SARS-CoV-2 severe acute respiratory syndrome coronavirus 2