SARS-CoV-2 infectivity by viral load, S gene variants and demographic factors and the utility of lateral flow devices to prevent transmission

Background: How SARS-CoV-2 infectivity varies with viral load is incompletely understood. Whether rapid point-of-care antigen lateral flow devices (LFDs) detect most potential transmission sources despite imperfect sensitivity is unknown. Methods: We combined SARS-CoV-2 testing and contact tracing data from England between 01-September-2020 and 28-February-2021. We used multivariable logistic regression to investigate relationships between PCR-confirmed infection in contacts of community-diagnosed cases and index case viral load, S gene target failure (proxy for B.1.1.7 infection), demographics, SARS-CoV-2 incidence, social deprivation, and contact event type. We used LFD performance to simulate the proportion of cases with a PCR-positive contact expected to be detected using one of four LFDs. Results: 231,498/2,474,066 (9%) contacts of 1,064,004 index cases tested PCR-positive. PCR-positive results in contacts independently increased with higher case viral loads (lower Ct values) e.g., 11.7%(95%CI 11.5-12.0%) at Ct=15 and 4.5%(4.4-4.6%) at Ct=30. B.1.1.7 infection increased PCR-positive results by ~50%, (e.g. 1.55-fold, 95%CI 1.49-1.61, at Ct=20). PCR-positive results were most common in household contacts (at Ct=20.1, 8.7%[95%CI 8.6-8.9%]), followed by household visitors (7.1%[6.8-7.3%]), contacts at events/activities (5.2%[4.9-5.4%]), work/education (4.6%[4.4-4.8%]), and least common after outdoor contact (2.9%[2.3-3.8%]). Contacts of children were the least likely to test positive, particularly following contact outdoors or at work/education. The most and least sensitive LFDs would detect 89.5%(89.4-89.6%) and 83.0%(82.8-83.1%) of cases with PCR-positive contacts respectively. Conclusions: SARS-CoV-2 infectivity varies by case viral load, contact event type, and age. Those with high viral loads are the most infectious. B.1.1.7 increased transmission by ~50%. The best performing LFDs detect most infectious cases.

The global health impact of SARS-CoV-2 is profound. 1 There is widespread on-going transmission 57 despite control efforts predominantly focused on quarantining symptomatic cases and population-58 level self-isolation. 2 The emergence of potentially more transmissible variants, such as B.1.1.7 3 which 59 has spread widely in the UK, has hampered control. However, vaccine roll-out offers the prospect of 60 reduced disease and transmission. 4 61 Intermittent national and regional social distancing and self-isolation measures have been imposed 62 in many countries. 5,6 Additional self-isolation measures for "contacts" (individuals exposed to SARS-63 CoV-2) vary by country, but generally last 7-14 days. 7 While reducing transmission, 64 quarantine/isolation measures have indirectly had many wider effects on economic productivity, 65 well-being 8 and non-COVID-19-related excess deaths. 9-11 Not all exposure to SARS-CoV-2 leads to 66 infection, e.g., in some settings only 5-7% of exposed "contacts" develop COVID-19 infection 12,13 and 67 modelling suggests ~15% of individuals are responsible for most SARS-CoV-2 transmission. 14 68 Therefore, using isolation selectively for those who are most infectious could lessen some of its 69 collateral impacts. 12,13 70 Our understanding of how individual infectiousness varies is limited. Several assays for infectivity 71 have been proposed. Functional assays include animal and cell culture models, whereas viral sub-72 genomic mRNA is a nucleic acid-based measure of infectivity. 15 Detection of viral protein, i.e. 73 antigen, as assessed by lateral flow devices (LFDs), has been shown to be more closely linked to viral 74 culture infectivity than PCR measurements. 16 However, few of these surrogate measures of 75 infectivity have been convincingly demonstrated to predict the real-world likelihood of a SARS-CoV-2 76 infected individual infecting someone else. 77 Here we use data from the England's national contact tracing and testing programs to explore the 78 relationship between infectivity and SARS-CoV-2 viral load, as measured by PCR cycle threshold (Ct)  79 values. We identify demographic factors associated with infectivity and assess the impact of the 80 emergence of the B.1.1.7 variant. We apply our results to a population of PCR-positive individuals to 81 estimate the proportion of infectious individuals detected by viral antigen LFDs under a range of 82 performance conditions. 83 84

85
Data from community and hospital PCR testing in England between 01-September-2020 and 28-86 February-2021 were obtained and linked with national contact tracing data by the UK Government 87 Department of Health and Social Care. Data extracts were de-identified prior to analysis and 88 included for PCR-confirmed cases and their contacts: demographic details (age, sex, ethnicity), if 89 symptoms were present for cases and the timing of testing relative to symptom onset, and test 90 results, as well as details on the nature of the contact events. 91 Index cases and contacts 92 We defined index cases as SARS-CoV-2 PCR-positive individuals with a community-based test 93 performed by three high-throughput national testing facilities ("Lighthouse Laboratories" in Milton 94 Keynes, Alderley Park or Glasgow), which reported Ct values indicating viral load. Samples were 95 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 5, 2021. were considered to identify positive results. 117 We used multivariable logistic regression to investigate associations between PCR-confirmed 118 infection in contacts (including contacts whether or not they had PCR tests) and the index case's Ct 119 value and SGTF status (B.1.1.7 proxy), the contact event nature, the case's demographics, and 120 incidence and social deprivation index at the contact's home location. We did not adjust for 121 symptoms in the case, as these may be mediators of the effect of viral load on onward transmission. 122 We used splines to account for non-linearity in continuous variables and screened for all pairwise 123 interactions between main effects (details in Supplement). 124 We performed sensitivity analyses to test our restriction to contacts tested 1-10 days after each 125 index case and including only contacts with PCR tests. We used unadjusted linear regression to 126 investigate the proportion of the variation in Ct values in contacts that could be explained by the 127 case's Ct value. 128 Simulations of the number of cases identified by antigen LFDs 129 We used our findings to estimate the proportion of potential transmission events where the source 130 case would have been detected using an antigen LFD, using existing data on the sensitivity of four 131 LFDs: Innova, Deep Blue, Orient gene and Abbott. 18 For each source case we simulated a positive or 132 negative LFD result by randomly drawing from the probability of a LFD being positive by the source 133 case's Ct value (see Supplement, Figure S1). Each simulation was repeated 1000 times. Additionally, 134 we ran simulations for a range of hypothetical LFD performances. 135 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 5, 2021. shared with other index cases, leaving 1,064,004 index cases in the analysis ( Figure 1, Table S1). The median (IQR) age of cases and contacts was 36 (24-51) and 31 (16-49) years respectively, and 160 54% and 52% with available data were female (Table 1). Most contact events occurred within 161 households (77.4%), followed by visits to households (8.3%), workplaces or education (8.0%), 162 attending events or activities (5.3%), and outdoors (0.3%). 163 Predictors of PCR-positive results in contacts 164 On univariable analysis ( Adjusted multivariable analysis showed strong evidence of effect modification (interactions) and 170 non-linear relationships, such that associations are best described graphically (Figures 2-3, S7-S11). 171 Index case Ct value was an important determinant of PCR-positive results in contacts, with an 172 approximately linear decline in positive results as Ct value increased, that was independent of the 173 nature of the contact event ( Figure 2). For example, amongst household contacts, with other 174 variables set to median values/baseline categories, rates of PCR-positive tests were 11.7% (95%CI 175 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 5, 2021. values near the upper limit of the assay, the relative increase in PCR-positive results fell to near 1 183 (Figure 3). In a sensitivity analysis restricted to contacts who had a PCR test (Table S2, Figures S13-S19), similar 198 relationships were seen between PCR-positive results and index case Ct values, contact type and 199 SGTF (Figures S19). While rates of PCR-positive results remained highest in older adult household 200 contacts, there was attenuation of the lower rates seen in children, consistent with main analysis 201 findings of less transmission from children arising from less testing being required or undertaken in 202 contacts of children ( Figure S18). In contrast to the main analysis, contacts of all non-white ethnic 203 groups (Table S2) and those living in more deprived areas ( Figure S17) were more likely to be PCR-204 positive, potentially due to differences in access to and use of testing by different ethnic and 205 socioeconomic groups. 206 Proportion of cases with PCR-positive contacts detected by LFDs  Figure S20). 211 As antigen LFD sensitivity varies by viral load, we used the distribution of viral loads in case-contact 212 pairs with a PCR-positive contact to simulate the proportion of such cases who would have been 213 detected using antigen LFDs ( Figure 5) higher relative infectiousness at moderate viral loads may represent increased infectiousness of 238 individual virions at viral loads where stochasticity is more important compared to higher viral loads. 239 The attenuation of the relative infectiousness at high Ct values partly arises from greater numbers of 240 wildtype strains exhibiting SGTF due to stochastic failure to detect a single gene at low viral loads. As 241 lower viral loads are less infectious, it may also reflect more PCR-positive contacts acquiring 242 infection from third parties, such that the characteristics of the index case matter less. This is 243 supported by the proportion of contacts testing PCR-positive not tending to zero at very low viral 244 loads. 245 85.4% PCR-positive contacts had an index case with a viral load of ≥10,000 RNA copies/ml (Ct≤24.4). 246 Hence, 85.4% of infections in contacts are potentially attributable to the 75.2% of cases overall with 247 a viral load of ≥10,000 RNA copies/ml. While such data could be used to drive differential 248 interventions to prevent onward transmission with a particular focus on those with high viral loads, 249 our findings suggest that most infected individuals still have some risk of transmitting onwards 250 based on Ct values. 251 However, we show that several LFDs are sufficiently sensitive to detect most cases that led to 252 onward transmission. These tests offer potential advantages, in returning a result in 15-30 minutes, 253 not requiring laboratory infrastructure and costing significantly less than PCR tests. However post-254 analytic infrastructure is still needed to collect results. Using the estimated sensitivity of four LFDs, 255 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 5, 2021. transmission. 294 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 5, 2021. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 5, 2021.   . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 5, 2021. listed. See Figure S11 for the multivariable relationship for deprivation score. 341 342 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 5, 2021. Role of the funding source 428 The funder of the study provided access to the data and facilitated data linkage. The funder had no 429 role in study design, data analysis, data interpretation, or writing of the report. The corresponding 430 author had full access to all the data in the study and had final responsibility for the decision to 431 submit for publication. 432 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 5, 2021. ; https://doi.org/10.1101/2021.03.31.21254687 doi: medRxiv preprint