COVID-19 due to the B.1.617.2 (Delta) variant compared to B.1.1.7 (Alpha) variant of SARS-CoV-2: a prospective observational cohort study

The Delta (B.1.617.2) variant was the predominant UK circulating SARS-CoV-2 strain between May and December 2021. How Delta infection compares with previous variants is unknown. This prospective observational cohort study assessed symptomatic adults participating in the app-based COVID Symptom Study who tested positive for SARS-CoV-2 from May 26 to July 1, 2021 (Delta overwhelmingly the predominant circulating UK variant), compared (1:1, age- and sex-matched) with individuals presenting from December 28, 2020 to May 6, 2021 (Alpha (B.1.1.7) the predominant variant). We assessed illness (symptoms, duration, presentation to hospital) during Alpha- and Delta-predominant timeframes; and transmission, reinfection, and vaccine effectiveness during the Delta-predominant period. 3581 individuals (aged 18 to 100 years) from each timeframe were assessed. The seven most frequent symptoms were common to both variants. Within the first 28 days of illness, some symptoms were more common with Delta versus Alpha infection (including fever, sore throat, and headache) and some vice versa (dyspnoea). Symptom burden in the first week was higher with Delta versus Alpha infection; however, the odds of any given symptom lasting ≥ 7 days was either lower or unchanged. Illness duration ≥ 28 days was lower with Delta versus Alpha infection, though unchanged in unvaccinated individuals. Hospitalisation for COVID-19 was unchanged. The Delta variant appeared more (1.49) transmissible than Alpha. Re-infections were low in all UK regions. Vaccination markedly reduced the risk of Delta infection (by 69-84%). We conclude that COVID-19 from Delta or Alpha infections is similar. The Delta variant is more transmissible than Alpha; however, current vaccines showed good efficacy against disease. This research framework can be useful for future comparisons with new emerging variants.

Illness profile. Symptoms within the first 28 days of illness are presented in Fig. 1
Correcting for age, sex, and vaccination status at time of positive test, and for false discovery rate, several symptoms were more common during the first 28 (Table 2), noting wide confidence intervals and variation over time. Estimates per region agreed broadly with some regional variation (Fig. 4, Table 2).  www.nature.com/scientificreports/ Effect of the Delta variant on re-infection. Figure 5 shows the (small) absolute numbers of re-infections across regions, with: a) the number of positive tests reported by app users; and b) the Delta variant as a proportion of circulating SARS-CoV-2, over time. Spearman correlations between reinfection and positive test incidence ranged from 0.46 in the South East to 0.83 in the Midlands. Correlations between reinfection and Delta variant proportions in each region were lower, ranging from 0.41 in the North East and Yorkshire to 0.69 in the North West. In most regions, the correlation of reinfections with the number of reported tests was higher than the correlation of reinfections with the proportion of Delta variant. Supplementary Table S5 presents characteristics of the bootstrapped distribution (100 samples) of correlations for each region over time. Thus, the rise of SARS-CoV-2 infection during the time of Delta predominance correlates more closely with the rise of incidence of new cases per se, rather than the rise in proportion of cases due to the Delta variant specifically.  (Fig. 6).

Discussion
Our large-scale community-based UK study has shown that COVID-19 is clinically similar whether due to Alpha or Delta variants. Ten of 31 symptoms were more common with Delta infection and one with Alpha infection. Although the burden of symptoms in the first week was higher with Delta infection, duration of many individual symptoms was shorter; fewer individuals experienced illness lasting more than 28 days-though saliently this was unchanged in unvaccinated individuals; and there was a trend towards fewer hospital presentations. These observations need to be interpreted in the context of increasing vaccination of the UK population, along with many other environmental and societal changes.
Few studies of COVID-19 due to the Delta variant are available for comparison. One study of 27 infected young individuals reported symptoms in 22 (81%), with the commonest symptoms fever (41%), cough (33%), headache (26%), and sore throat (26%) (duration of illness not reported) 18. Other studies report smaller cohorts. The REACT-1 study Round 14 report (UK data during September 2021, with Delta variant the predominant UK variant) showed a weighted prevalence of individuals testing positive varied greatly by age (0.29% in adults aged > 75 years compared to 2.55% in teenagers and 2.32% in children aged 5-12 years), noting high vaccination rates in older individuals and little or no vaccination in younger age groups at the time of this report 19 . However, data on symptoms (duration and/or prevalence) were not reported.
The risk of LC28 was lower with Delta (8.7%) versus Alpha infection (10.6%), although not statistically different in unvaccinated individuals. These results are similar to our previous paper using similar methodology, for individuals infected during the first UK pandemic wave (13.3%) 17 . The Post COVID syndrome (Long COVID) is defined as illness duration > 12 weeks after likely SARS-CoV-2 infection (https:// www. nice. org. uk/ guida nce/ NG188). Our census dates preclude our ability to compare illness duration beyond 28 days between our two cohorts. Estimates of prevalence of the Post-COVID syndrome are difficult, as many studies lack appropriate control groups. A recent meta-analysis of UK longitudinal cohort studies suggested the post-COVID syndrome was present in 1.2-4.8% of individuals 20 , similar to recently published figures from the Office for National Statistics (ONS) (1.9% of the UK population self-reporting long COVID (diagnosis otherwise unverified) as of October 2, 2021, although only 71% had (or suspected they had) COVID-19 12 weeks earlier (https:// www. ons. gov. uk/ peopl epopu latio nandc ommun ity/ healt hands ocial care/ condi tions anddi seases/ bulle tins/ preva lence ofong oings ympto msfol lowin gcoro navir uscov id19i nfect ionin theuk/ 4nove mber2 021).
We showed a marked increase in transmissibility with the Delta versus non-Delta (i.e., Alpha) variant, noting wide confidence intervals. This analysis does not take into account prior natural infection or vaccination rates within the community; and is likely a combination of both the Delta variant's transmission advantage and its potential ability to evade immunity (whether induced by vaccination or prior natural infection with Alpha or other strains). This estimated increase in transmissibility is greater than we previously estimated for Alpha versus earlier variants using the same methodology [1.35 (95% CI 1.02-1.69)] 15 noting again pertinent differences (e.g., viral prevalence, lockdown restrictions) between the current and previous studies. Estimates in both studies assume that incidence estimated from app users can be made representative of the wider population, using stratification by age and vaccination status. However, other factors such as behavior and socio-economic status are not corrected for by this analysis. Other studies have also identified higher Delta transmissibility, resulting in rising incidence particularly in young unvaccinated age groups, higher re-infection rates, and a higher viral load in infected individuals 21 . Here we note the REACT-1 study report of an exponential increase in infections in children aged 5-17 years in September 2021, coinciding with return-to-school, with most school-age children unvaccinated at this time.
Our study found that for most regions of the UK, the correlation of reinfections with the number of reported positive tests (i.e., incidence of cases) was higher than the correlation of reinfections with the proportion of Delta www.nature.com/scientificreports/ among the circulating variants (i.e., incidence of variant). In other words, the rise in COVID-19 correlated more closely with increase in prevalence of SARS-CoV-2 overall rather than the increased proportion of circulating SARS-CoV-2 due to the Delta variant. SARS-CoV-2 infection provides substantial and persistent immunologic protection for at least several months for most individuals, with a recent systematic review suggesting a risk reduction of reinfection of > 90%, similar to vaccination, and evident for at least 10 months 22 . However, this may not be uniform across the population. A study of tested individuals followed prospectively for at least 3 months Odds ratios for any symptom presenting within the first 28 days of illness in individuals with COVID-19 during periods of SARS-CoV-2 Delta versus Alpha variant predominance. Age, gender, and vaccination.status are included as covariates in this analysis. Red markers encode statistical significance with α-value < 0.05, whereas grey markers encode non-significant differences, after correction for false discovery rate. Our observational data support effectiveness of both BNT162b2 and ChAdOx1 nCoV-19 vaccines against the Delta variant. Both reduced the risk of testing positive during the Delta period, evident after the first and enhanced after the second dose 25 . These figures are similar to our previous results when the Alpha variant was predominant 15 . We have an inherent bias due to the nature of the UK vaccine rollout, whereby health-care workers, elderly people, and clinically vulnerable individuals were prioritised before the younger population, creating unbalanced demographic characteristics between vaccinated and unvaccinated populations. Moreover, we cannot compare vaccination effectiveness against Alpha versus Delta variants, given the many differences between the two timeframes. Although we attempted to adjust for some of these differences using Poisson regression, behavioural factors are difficult to capture. For example, individuals vaccinated earlier may have changed their behaviours over concern of possible waning antibody status and possible reducing immunity 26 . However, our results concord with vaccination trial data 27,28 , and provide support for ongoing vaccination campaigns internationally. Previous data have shown that vaccination is associated with significant reduction in risk of hospitalisation and disease progression to death or mechanical ventilation in individuals with COVID-19 12,29 . Our data similarly showed a trend towards fewer hospital presentations. Later analyses during further waves of the pandemic will be useful here with the methods and approaches described here similarly applied to assess the impact of emerging variants.
We acknowledge the limitations of our observational study. Self-reported data from a mobile phone app may disproportionately represent more affluent populations and can introduce information bias and/or effect bias, although previous work from the CSS has shown that our self-reported data aligns well with surveys designed to be representative of the population 30 and smartphone ownership in the UK is extensive with little evidence that this varies greatly across socio-economic groups. Further, 1600 of 7162 individuals were proxy-reported, which may also affect symptom reporting, although the proportion of proxy-reported individuals as a percentage of total individuals was the same for both time-periods (22.4%). Participants could only report a positive test and we cannot confirm the actual variant causing infection, although our assumptions of Delta and Alpha infection are strongly supported by UK-COG surveillance variant testing. During the study, both overall numbers and individual app users fluctuated in their participation in the CSS-app, potentially for many factors including mass-media information, summer vacation, and perception of relevance. Our populations were matched for age and sex but not BMI; and we note higher diabetes prevalence and BMI in the Alpha cohort. Relevantly, vaccination was not only tiered by age but also to those with co-morbidities including diabetes. As mentioned, the timeframes of Alpha and Delta variant predominance differed with respect to guidance on social distancing and behavior in public spaces, highly likely to affect viral diffusion in the population, thus affecting our transmission calculations. Last, vaccine effectiveness could only be determined in tested individuals, noting that we do not have information regarding the reason for testing in these individuals. We were also only able to assess individuals in the age range of 20-65, in order to avoid unbalanced case/control data. Relevantly, early post-vaccination symptoms can mimic COVID-19 31 but may not necessarily trigger testing. Here, our previous work showed that vaccinated individuals are more likely to have post-vaccination systemic symptoms after a previously positive test compared to those without known past infection (odds ratios 2.3-4.0), which may bias presentation for SARS-CoV-2 testing post-vaccination.  Supplementary  Information). Briefly, upon enrolment users provide baseline demographic and health information, and subsequently are prompted daily to record symptoms (or their absence) through direct questioning (Supplementary  Table S2) and free text, any SARS-CoV-2 testing and corresponding result, vaccination details, and any hospital presentation. Users can also proxy-report for others. The current study was drawn from approximately 1 million  www.nature.com/scientificreports/ UK app users who logged data at least once between December 28, 2020 to July 1, 2021. Data were extracted and curated through ExeTera software 33 . Ethics approval was granted by the KCL Ethics Committee (LRS-19/20-18210). To ensure informed consent, at registration with the app, all participants were shown informative documentation, and were offered to provide consent for their data to be used for COVID-19 research. Governance was specifically granted for use of proxy-reported data. Research was performed in accordance with the relevant guidelines and regulations, and particularly in full compliance with the Declaration of Helsinki and further updates.
Data from all UK adult participants aged 18 to 100 years (including proxy-reported individuals) who logged a positive PCR or lateral flow antigen test (LFAT) for SARS-CoV-2 between December 28, 2020 to July 1, 2021 were considered. As previously 17 , individuals were considered to have COVID-19 if SARS-CoV-2-associated symptoms were reported (or proxy-reported) (Supplementary Table S2) between two weeks before and one week after positive testing. Data were included for individuals who reported at least weekly, from first symptom report until returning to symptom-freedom or until reporting ceased 22 .
Data were compared between two time periods: December 28, 2020 to May 6, 2021, when the Alpha variant was the predominant circulating SARS-CoV-2 strain (proportion of sequenced strains: > 75% from December 28, 2020, reaching > 95% by February 3 2021, and remaining > 75% until 28 April); and May 26, 2021 to July 1, 2021, when the Delta variant was the predominant strain (> 75% from May 26, reaching > 95% by June 9 and > 99% from June 30 to data census date) (Supplementary Table S1). Individuals logging a positive test did not have variant confirmation by sequencing; thus, illness within these two timeframes was attributed to the predominant circulating variant. Terminology herein reflects this assumption.
Through an Euclidean distance-based algorithm 34 , individuals with Delta infection were matched 1:1, based on their age and sex, with individuals with Alpha infection. We were unable to match for SARS-CoV-2 prevalence, tiered lockdown restrictions, or vaccination rates, which varied widely across the community and with time during this study.
Symptom data were censored at August 5, 2021, 35 days after last inclusion date for testing positive with Delta infection, allowing at least 28 days' symptom evaluation for all individuals. Symptoms were considered over the entire illness, which by virtue of illness definition could extend outside SARS-CoV-2 testing date boundaries (a maximum of two weeks before and five weeks after testing, allowing for individuals whose illness started up to a week after positive test). To allow for symptom waxing and waning, individuals who returned a healthy report but subsequently logged as symptomatic within seven days of their last unhealthy report were considered unwell from their initial illness, with per-symptom and illness duration calculated accordingly.
We ascertained odds of a given symptom developing within 28 days of illness; and odds of each symptom lasting ≥ 7 days, corrected for age, sex and vaccination status (unvaccinated, 1 dose, 2 doses), with a given vaccination considered valid after 14 days (allowing for evolving immunity). We used false discovery rate to account for multiple comparisons. We assessed risk of illness duration ≥ 28 days (LC28) and hospital presentation (admission www.nature.com/scientificreports/ or emergency room attendance), in the cohort overall (similarly adjusted for age, sex and vaccination status) and in unvaccinated individuals alone, as a sensitivity analysis.
Transmissibility. We used data from COVID-19 Genomics UK Consortium (COG-UK) to extract timeseries of the percentage of daily positive SARS-CoV-2 testing from the Delta lineage in Scotland, Wales, and each of nine National Health Service (NHS) regions in England. Northern Ireland was excluded due to low sample numbers in the COG-UK dataset. The COG-UK data are produced by sequencing a random sample of positive PCR tests from the general community. Daily SARS-CoV-2 incidence data for Scotland, Wales, and each NHS region in England were estimated from March 14 to August 8, 2021, using CSS app data and previously described methodology 35 . The method uses both positive SARS-CoV-2 test results and symptom reports from app users, to estimate incidence. Data are stratified by age and vaccination status to ensure estimates made from the CSS app population are representative of the wider population.
Using COG-UK data to estimate the proportion of Delta in circulation in each region per day, incidence estimates were decomposed into two incidence time-series per region, one for 'non-Delta' (in the timeframe considered here, predominantly Alpha [Supplementary Table S1]) and one for Delta, assuming that the two incidence time-series should sum to match total incidence. R(t) was estimated separately for non-Delta and Delta variants, using previously described methodology 35 . Briefly, we used the relationship I t+1 = I t exp(μ (R(t) -1)), where 1/μ is the serial interval and I t the incidence on day t. We modelled the system as a Poisson process and assumed the serial interval was drawn from a gamma distribution with α = 6.0 and β = 1.5; and used Markov Chain Monte-Carlo to estimate R(t). We compared both multiplicative and additive differences of the new and old R(t) values for days when the Delta proportion in a region was > 3%.
Reinfection during rise of Delta variant. Reinfection was defined as previously 36 (presence of two positive PCR or LFAT tests separated by > 90 days, with an asymptomatic period of ≥ 7 days before the second positive test). To assess risk of reinfection during the Delta variant timeframe we performed ecological studies for each region, examining the Spearman correlation between the proportion of circulating SARS-CoV-2 due to Delta (Supplementary Table S5, Supplementary Figure S1) and number of reinfections per week over time, assessed from 10 weeks prior to Delta prevalence of 25% until 10 weeks after Delta prevalence of 75% (22 weeks); and between the number of positive tests reported through the app and the number of reinfections. We compared the bootstrapped distributions of these two correlations in each region, using the Mann-Whitney U test (Supplementary Table S5).

Post-vaccination infection during Delta period.
We analysed 515,138 app users who reported vaccination with BNT162b2 (BioNTech-Pfizer) or ChAdOx1 nCoV-19 (Oxford-Astra Zeneca) and were subsequently tested for SARS-CoV-2 (PCR or LFAT) 14-60 days after either first or second vaccination (assessed separately) after 26 May 2021 36 . Age was restricted to 20-65 years, as most individuals > 65 years were vaccinated and most individuals < 20 years unvaccinated during the time of analysis, biasing the control groups for these ages. Users who had reported SARS-CoV-2 infection previously were excluded. Unvaccinated users reporting SARS-CoV-2 test results in the same or following week as a vaccinated app user served as controls. In the event of multiple tests logged for an individual vaccinated user, either the first positive or the last negative result was selected. For each vaccine and per dose, we modelled rates of positive testing in vaccinated versus unvaccinated individuals, using Poisson regressions adjusting for number of tests, age, co-morbidities, sex, healthcare worker status, obesity, and weekly incidence in the community (by controlling for the date of the test). The adjusted risk reduction was then calculated as RR = riskratio i,n − 1, where i is the vaccine type, and riskratio is the ratio of infection rates in vaccinated individuals compared to unvaccinated individuals, derived from our Poisson model.

Data availability
Data collected in the COVID Symptom Study smartphone app can be shared with other health researchers through the UK National Health Service-funded Health Data Research UK and Secure Anonymised Information Linkage consortium, housed in the UK Secure Research Platform (Swansea, UK). Anonymised data are available to be shared with researchers according to their protocols in the public interest, and an administrative fee might apply. Link at https:// web. www. healt hdata gatew ay. org/ datas et/ 594cf e55-96e3-45ff-874c-2c000 6eeb8 81. Information on anonymized SARS-CoV-2 test results is recorded as positive/negative/invalid, and sequencing data were not captured.