SummaryBackgroundBy August, 2021, South Africa had been affected by three waves of SARS-CoV-2; the second associated with the beta variant and the third with the delta variant. Data on SARS-CoV-2 burden, transmission, and asymptomatic infections from Africa are scarce. We aimed to evaluate SARS-CoV-2 burden and transmission in one rural and one urban community in South Africa.MethodsWe conducted a prospective cohort study of households in Agincourt, Mpumalanga province (rural site) and Klerksdorp, North West province (urban site) from July, 2020 to August, 2021. We randomly selected households for the rural site from a health and sociodemographic surveillance system and for the urban site using GPS coordinates. Households with more than two members and where at least 75% of members consented to participate were eligible. Midturbinate nasal swabs were collected twice a week from household members irrespective of symptoms and tested for SARS-CoV-2 using real-time RT-PCR (RT-rtPCR). Serum was collected every 2 months and tested for anti-SARS-CoV-2 antibodies. Main outcomes were the cumulative incidence of SARS-CoV-2 infection, frequency of reinfection, symptomatic fraction (percent of infected individuals with ≥1 symptom), the duration of viral RNA shedding (number of days of SARS-CoV-2 RT-rtPCR positivity), and the household cumulative infection risk (HCIR; number of infected household contacts divided by the number of susceptible household members).Findings222 households (114 at the rural site and 108 at the urban site), and 1200 household members (643 at the rural site and 557 at the urban site) were included in the analysis. For 115 759 nasal specimens from 1200 household members (follow-up 92·5%), 1976 (1·7%) were SARS-CoV-2-positive on RT-rtPCR. By RT-rtPCR and serology combined, 749 of 1200 individuals (62·4% [95% CI 58·1–66·4]) had at least one SARS-CoV-2 infection episode, and 87 of 749 (11·6% [9·4–14·2]) were reinfected. The mean infection episode duration was 11·6 days (SD 9·0; range 4–137). Of 662 RT-rtPCR-confirmed episodes (>14 days after the start of follow-up) with available data, 97 (14·7% [11·9–17·9]) were symptomatic with at least one symptom (in individuals aged <19 years, 28 [7·5%] of 373 episodes symptomatic; in individuals aged ≥19 years, 69 [23·9%] of 289 episodes symptomatic). Among 222 households, 200 (90·1% [85·3–93·7]) had at least one SARS-CoV-2-positive individual on RT-rtPCR or serology. HCIR overall was 23·9% (195 of 817 susceptible household members infected [95% CI 19·8–28·4]). HCIR was 23·3% (20 of 86) for symptomatic index cases and 23·9% (175 of 731) for asymptomatic index cases (univariate odds ratio [OR] 1·0 [95% CI 0·5–2·0]). On multivariable analysis, accounting for age and sex, low minimum cycle threshold value (≤30 vs >30) of the index case (OR 5·3 [2·3–12·4]) and beta and delta variant infection (vs Wuhan-Hu-1, OR 3·3 [1·4–8·2] and 10·4 [4·1–26·7], respectively) were associated with increased HCIR. People living with HIV who were not virally supressed (≥400 viral load copies per mL) were more likely to develop symptomatic illness when infected with SAR-CoV-2 (OR 3·3 [1·3–8·4]), and shed SARS-CoV-2 for longer (hazard ratio 0·4 [95% CI 0·3–0·6]) compared with HIV-uninfected individuals.InterpretationIn this study, 565 (85·3%) SARS-CoV-2 infections were asymptomatic and index case symptom status did not affect HCIR, suggesting a limited role for control measures targeting symptomatic individuals. Increased household transmission of beta and delta variants was likely to have contributed to successive waves of SARS-CoV-2 infection, with more than 60% of individuals infected by the end of follow-up.FundingUS CDC, South Africa National Institute for Communicable Diseases, and Wellcome Trust.