background: Large-scale vaccination against COVID-19 is being implemented in many countries with CoronaVac, an inactivated vaccine. We aimed to assess the immune persistence of a two-dose schedule of CoronaVac, and the immunogenicity and safety of a third dose of CoronaVac, in healthy adults aged 18 years and older. methods: In the first of two single-centre, double-blind, randomised, placebo-controlled phase 2 clinical trials, adults aged 18-59 years in Jiangsu, China, were initially allocated (1:1) into two vaccination schedule cohorts: a day 0 and day 14 vaccination cohort (cohort 1) and a day 0 and day 28 vaccination cohort (cohort 2); each cohort was randomly assigned (2:2:1) to either a 3 μg dose or 6 μg dose of CoronaVac or a placebo group. Following a protocol amendment on Dec 25, 2020, half of the participants in each cohort were allocated to receive an additional dose 28 days (window period 30 days) after the second dose, and the other half were allocated to receive a third dose 6 months (window period 60 days) after the second dose. In the other phase 2 trial, in Hebei, China, participants aged 60 years and older were assigned sequentially to receive three injections of either 1·5 μg, 3 μg, or 6 μg of vaccine or placebo, administered 28 days apart for the first two doses and 6 months (window period 90 days) apart for doses two and three. The main outcomes of the study were geometric mean titres (GMTs), geometric mean increases (GMIs), and seropositivity of neutralising antibody to SARS-CoV-2 (virus strain SARS-CoV-2/human/CHN/CN1/2020, GenBank accession number MT407649.1), as analysed in the per-protocol population (all participants who completed their assigned third dose). Our reporting is focused on the 3 μg groups, since 3 μg is the licensed formulation. The trials are registered with ClinicalTrials.gov, NCT04352608 and NCT04383574. results: 540 (90%) of 600 participants aged 18-59 years were eligible to receive a third dose, of whom 269 (50%) received the primary third dose 2 months after the second dose (cohorts 1a-14d-2m and 2a-28d-2m) and 271 (50%) received a booster dose 8 months after the second dose (cohorts 1b-14d-8m and 2b-28d-8m). In the 3 μg group, neutralising antibody titres induced by the first two doses declined after 6 months to near or below the seropositive cutoff (GMT of 8) for cohort 1b-14d-8m (n=53; GMT 3·9 [95% CI 3·1-5·0]) and for cohort 2b-28d-8m (n=49; 6·8 [5·2-8·8]). When a booster dose was given 8 months after a second dose, GMTs assessed 14 days later increased to 137·9 (95% CI 99·9-190·4) for cohort 1b-14d-8m and 143·1 (110·8-184·7) 28 days later for cohort 2b-28d-8m. GMTs moderately increased following a primary third dose, from 21·8 (95% CI 17·3-27·6) on day 28 after the second dose to 45·8 (35·7-58·9) on day 28 after the third dose in cohort 1a-14d-2m (n=54), and from 38·1 (28·4-51·1) to 49·7 (39·9-61·9) in cohort 2a-28d-2m (n=53). GMTs had decayed to near the positive threshold by 6 months after the third dose: GMT 9·2 (95% CI 7·1-12·0) in cohort 1a-14d-2m and 10·0 (7·3-13·7) in cohort 2a-28d-2m. Similarly, in adults aged 60 years and older who received booster doses (303 [87%] of 350 participants were eligible to receive a third dose), neutralising antibody titres had declined to near or below the seropositive threshold by 6 months after the primary two-dose series. A third dose given 8 months after the second dose significantly increased neutralising antibody concentrations: GMTs increased from 42·9 (95% CI 31·0-59·4) on day 28 after the second dose to 158·5 (96·6-259·2) on day 28 following the third dose (n=29). All adverse reactions reported within 28 days after a third dose were of grade 1 or 2 severity in all vaccination cohorts. There were three serious adverse events (2%) reported by the 150 participants in cohort 1a-14d-2m, four (3%) by 150 participants from cohort 1b-14d-8m, one (1%) by 150 participants in each of cohorts 2a-28d-2m and 2b-28d-8m, and 24 (7%) by 349 participants from cohort 3-28d-8m. conclusions: A third dose of CoronaVac in adults administered 8 months after a second dose effectively recalled specific immune responses to SARS-CoV-2, which had declined substantially 6 months after two doses of CoronaVac, resulting in a remarkable increase in the concentration of antibodies and indicating that a two-dose schedule generates good immune memory, and a primary third dose given 2 months after the second dose induced slightly higher antibody titres than the primary two doses. background: National Key Research and Development Program, Beijing Science and Technology Program, and Key Program of the National Natural Science Foundation of China. unassigned: For the Mandarin translation of the abstract see Supplementary Materials section.