This single-center, randomized, prospective, open-labelled, pilot cohort study enrolled healthy adults aged 18-60 years during September to December of 2021. Participants naïve for SARS-CoV-2 infection, capable of adhering to scheduled fractional dosing regimens, with an ability to understand Thai, and to self-report digitally were included in the study. Exclusion criteria included those that: were previously infected with SARS-CoV-2; received prophylactic or investigational COVID-19 treatment; received blood, plasma, immunoglobulins, or antibodies within 90 days of the study; had a history of severe drug or vaccine allergies, pre-existing comorbidities, underlying diseases, drug, or substance abuse; were pregnant; and were immunocompromised or receiving immunosuppressive agents. These study protocols were approved by the Human Research Protection Unit, Faculty of Medicine Siriraj Hospital, Mahidol University (COA: MU-MOU 704/2021) and registered under Thailand’s Clinical Trial Registry (registration number: TCTR20210904004, https://www.thaiclinicaltrials.org/). All participants provided written informed consent.

Initially, 10 participants were randomly assigned into 4 groups each to receive two homologous or heterologous ID doses of CoronaVac (Sinovac), ChAdOx1 nCoV-19 (AstraZeneca), or BNT162b2 (Pfizer-BioNTech) 7 days apart. The vaccine regimens (first-second dose) for each arm were: BNT162b2-BNT162b2 (Group 1), ChAdOx1-BNT162b2 (Group 2), CoronaVac-ChAdOx1 (Group 3), and ChAdOx1-ChAdOx1 (Group 4). Participants received two injections of each vaccination, one in each arm within the deltoid region. The dosage for each injection was 10-20% of conventional IM doses based on previously published literature (3). This entailed 0.1 mL for CoronaVac and ChAdOx1 (20% of standard 0.5 mL IM dosage) and 0.05 mL for BNT162b2 (17% of standard 0.3 mL IM dosage). Blood samples were collected before each dose and 2 weeks after the second dose for immunologic evaluations, including: anti-receptor binding domain of Wuhan strain S1-subunit spike IgG (anti-RBD IgG), neutralizing antibodies (NAb) for Wuhan, and cellular immune responses against Wuhan. As an exploratory outcome, NAb levels were also measured against omicron subvariants. Baseline samples were also tested for SARS-CoV-2 anti-nucleocapsid protein (anti-NP) antibodies to determine prior natural infection. Participants across all four regimens were instructed to digitally self-assess and report their solicited local and/or systemic AEs after each dose. Solicited local AEs entailed injection site reactions, while systemic AEs included: myalgia, fatigue, headache, fever, diarrhea, and nausea. Both types of AEs were graded numerically on scales of 1 to 4 based on the Common Terminology Criteria for Adverse Events (Version 5.0) by the United States National Cancer Institute (NCI/NIH). Mild scores (1) did not interfere with patients’ activities, moderate scores (2) interfered with patients’ activity, severe scores (3) prevented daily activities, and emergency scores (4) required hospitalization 4 (12).

Preliminary analyses of anti-RBD IgG were performed for all regimens and participants. Evidence of positive anti-NP or anti-RBG IgG at baseline were excluded. In the extended phase, additional participants were recruited to the two groups (10 per regimen) with the highest anti-RBD IgG measured 2 weeks after the second vaccination to meet statistical power. Additional blood sample were collected 12 weeks after the second dose for these two groups to evaluate the persistence of anti-RBG IgG.

Plasma samples were isolated using sodium citrate and stored at -80°C. A chemiluminescent microparticle assay (CMIA) was used to determine anti-RBD and anti-NP through SARS-CoV-2 IgG II Quant (Abbott Laboratory System, Illinois, US) on the ARCHITECT i System. Antibody levels were linearly measured between 21.0-40,000.0 arbitrary units per mL (AU/mL), and subsequently converted to binding antibody units per mL (BAU/mL) per WHO’s International Standards and an equation provided by the manufacturer (BAU/mL = 0.142 × AU/mL). Seropositivity was defined by cutoff values ≥ 50 AU/mL (7.1 BAU/mL). Antibody response against SARS-CoV-2 NP protein was determined using baseline plasma samples through CMIA SARS-CoV-2 IgG (Abbott, List No. 06R86) on the ARCHITECT i System.

PVNT was carried out as described previously at the National Center of Genetic Engineering and Biotechnology, Thailand (13). Assays were performed against the ancestral Wuhan strain, and omicron subvariants BA.1, BA.2, and BA.5. Antibody titers (PVNT₅₀) were calculated using GraphPad Prism 9 version 9.2.0 (GraphPad Software, CA, USA) to interpolate the point at which pseudovirus infectivity had been reduced by 50% of the value found for no serum control samples. The limit of detection (LOD) was 1:40.

T-cell responses were assessed using human interferon-gamma (IFN-γ) ELISpot kits according to the manufacturers’ instruction (Mabtech AB, Nacka Strand, Sweden – as previously described (14)). This entailed the use of two peptide pools : (1) an S-defined peptide pool (Mabtech) consisting of 100 peptides from spike (S) protein (purity of 90%); and (2) an NMO-defined peptide pool (Mabtech) consisting of 101 peptides from nucleocapsid (N), membrane (M), open reading frame (ORF) 1, non-structural protein (nsp) 3, ORF-3a, ORF-7a, and ORF-8 proteins (purity of 87%). The ELISpot plates were read using IRIS (Mabtech) and spots were analyzed using Apex software 1.1 (Mabtech) and converted to spot-forming units (SFU) per million cells. The cut-off for positive response was set as 20 SFU.

Participants positive for anti-NP and anti-RBD at baseline were excluded from the analysis. Immunological endpoints (anti-SARS-CoV-2 RBD IgG, PVNT₅₀, and ELISpot SFU) were reported as geometric means (GMs), with 95% confidence intervals (CI). PVNT₅₀ titers that were lower than LOD of 40 were assigned a value of 20. As all participants were seronegative at baseline, seroconversion was determined as becoming seropositive (anti-RBD IgG ≥ 50 AU/mL, PVNT₅₀ ≥ 40) two weeks after the second vaccination. Sample sizes of 20 per regimen would provide the ability to detect differences between ID and conventional IM regimens with 85% confidence. GraphPad Prism 9 version 9.2.0 (GraphPad Software, CA, USA) was used to perform unpaired t-tests for GM comparisons between treatment regimens. ANOVA and STATA version 17 (StataCorp, LP, College Station, TX, USA) were used to examine GMs for different regimens. Endpoints of AEs were presented as frequencies and Chi-square tests were used to test for statistical differences between endpoints. We represented this information as GMs of anti-RBD-IgG and AE frequency, similar to another report of ours regarding IM regimens in healthy volunteers (15).

None of the participants were seropositive for anti-RBD IgG 7 days after the first dose. All were seropositive after the second dose. Anti-SARS-CoV-2 RBD IgG geometric mean concentrations (GMCs) 2 weeks after the second dose had significantly increased (P < 0.001) compared to baseline values for all four regimens: 414.84 (95% confidence interval (CI): 316.96, 542.96) BAU/mL for Group 1, 597.29 (95% CI: 411.37, 867.25) BAU/mL for Group 2, 73.51 (95% CI: 44.09, 122.57) BAU/mL for Group 3, and 138.56 (95% CI: 43.59-440.46) BAU/mL for Group 4 (see Figure 2A and Supplementary Table 1 ). No statistical significance in anti-RBD IgG was observed 2 weeks after the second vaccination as seen between Groups 1 and 2, as well as Groups 3 and 4. However, Groups 1 and 2 induced significantly higher anti-RBD IgG than Groups 3 and 4 (P < 0.001 for each comparison). 2 weeks after second dose, each group had significantly lower anti-RBD IgG (5.4, 3.6, 11.6, and 2.0 times lower for Groups 1 to 4, respectively) compared to reference anti-RBD IgG values from conventional IM regimens (P < 0.001 for each comparison, except Group 4). Anti-RBD IgG for Groups 1 and 2 decreased by 3.1 and 3.8 times, respectively, 12 weeks following the second dose.

The PVNT₅₀ geometric mean titer (GMT) and seroconversion against the ancestral Wuhan strain 2 weeks after the second vaccination are shown in Figure 2B and Supplementary Table 1 . Seroconversion rates varied between regimens. Higher seroconversion rates (75% and 90% for Groups 1 and 2, respectively) were observed in those vaccinated with at least one BNT162b2 dose compared to those with CoronaVac-ChAdOx1 or two-doses of ChAdOx1 (57.1% and 37.5% for Groups 3 and 4, respectively, as shown in Figure 2B ). The exploratory outcome for the seroconversion rate and PVNT₅₀ titers against omicron subvariants after two doses was low or undetectable ( Supplementary Figures 1A, B , and Supplementary Table 1 ). These low PVNT₅₀ responses against omicron were like those generated in their respective IM vaccine regimens.

All participants, except one, had negative responses at baseline. One participant had a low ELISpot response to S-protein (24 SFU). 2 weeks after the second dose, all participants across the four regimens had significant increases in IFN-γ response against S-protein. The highest GM SFU for S-protein 2 weeks after the second dose was observed in Group 2 (441.34; 95% CI 271.10, 718.47), followed by Group 1 (373.80; 95% CI 246.12, 567.72), Group 4 (224.99; 95% CI 136.63, 370.50), and Group 3 (85.88; 95% CI 42.22, 174.70) (see Figure 3 and Supplementary Table 1 ). There were no significant differences in IFN-γ responses against S-protein between Groups 1, 2, and 4. Group 3 had a significantly lower response compared to the other three groups (P = 0.002) (see Supplementary Table 1 ). There was no significant increase in IFN-γ response against NMO proteins for all four groups (see Figure 3 and Supplementary Table 1 ).

Many reported AEs were mild, some moderate, but none severe. All AEs fully resolved before the end of the study (see Figures 4A, B , Supplementary Table 2 , and Supplementary Figure 2 ). Compared to the second IM dose, a lower proportion of systemic AEs were reported after both ID doses for each respective vaccine regimen. The only exception was homologous ChAdOx1 (Group 4), with similar systemic AEs following ID or IM administration (see Figure 4A ). Compared to IM injection, a lower proportion of local AEs were reported for Groups 1 and 2, but not after ChAdOx1 administration in Groups 3 and 4.