Intended for healthcare professionals

CCBYNC Open access

Effectiveness of heterologous and homologous covid-19 vaccine regimens: living systematic review with network meta-analysis

BMJ 2022; 377 doi: (Published 31 May 2022) Cite this as: BMJ 2022;377:e069989
  1. Wing Ying Au, masters student1 2,
  2. Peter Pak-Hang Cheung, assistant professor1 2
  1. 1Li Ka Shing Institute of Health Sciences, Chinese University of Hong Kong, Hong Kong, China
  2. 2Department of Chemical Pathology, Chinese University of Hong Kong, Hong Kong, China
  1. Correspondence to: P P-H Cheung ppcheung{at}
  • Accepted 27 April 2022


Objective To evaluate the effectiveness of heterologous and homologous covid-19 vaccine regimens with and without boosting in preventing covid-19 related infection, hospital admission, and death.

Design Living systematic review and network meta-analysis.

Data sources World Health Organization covid-19 databases, including 38 sources of published studies and preprints.

Study selection Randomised controlled trials, cohort studies, and case-control studies.

Methods 38 WHO covid-19 databases were searched on a weekly basis from 8 March 2022. Studies that assessed the effectiveness of heterologous and homologous covid-19 vaccine regimens with or without a booster were identified. Studies were eligible when they reported the number of documented, symptomatic, severe covid-19 infections, covid-19 related hospital admissions, or covid-19 related deaths among populations that were vaccinated and unvaccinated. The primary measure was vaccine effectiveness calculated as 1−odds ratio. Secondary measures were surface under the cumulative ranking curve (SUCRA) scores and the relative effects for pairwise comparisons. The risk of bias was evaluated by using the risk of bias in non-randomised studies of interventions (ROBINS-I) tool for all cohort and case-control studies. The Cochrane risk of bias tool (version 2; ROB-2) was used to assess randomised controlled trials.

Results The first round of the analysis comprised 53 studies. 24 combinations of covid-19 vaccine regimens were identified, of which a three dose mRNA regimen was found to be the most effective against asymptomatic and symptomatic covid-19 infections (vaccine effectiveness 96%, 95% credible interval 72% to 99%). Heterologous boosting using two dose adenovirus vector vaccines with one mRNA vaccine has a satisfactory vaccine effectiveness of 88% (59% to 97%). A homologous two dose mRNA regimen has a vaccine effectiveness of 99% (79% to 100%) in the prevention of severe covid-19 infections. Three dose mRNA is the most effective in reducing covid-19 related hospital admission (95%, 90% to 97%). The vaccine effectiveness against death in people who received three doses of mRNA vaccine remains uncertain owing to confounders. In the subgroup analyses, a three dose regimen is similarly effective in all age groups, even in the older population (≥65 years). A three dose mRNA regimen works comparably well in patients who are immunocompromised and those who are non-immunocompromised. Homologous and heterologous three dose regimens are effective in preventing infection by covid-19 variants (alpha, delta, and omicron).

Conclusion An mRNA booster is recommended to supplement any primary vaccine course. Heterologous and homologous three dose regimens work comparably well in preventing covid-19 infections, even against different variants. The effectiveness of three dose vaccine regimens against covid-19 related death remains uncertain.

Systematic review registration This review was not registered. The protocol is included in the supplementary document.

Readers’ note This article is a living systematic review that will be updated to reflect emerging evidence. Updates may occur for up to two years from the date of original publication.


The covid-19 pandemic caused by SARS-CoV-2 has led to more than 489 million confirmed cases and six million deaths worldwide according to the World Health Organization covid-19 weekly epidemiological update on 5 April 2022.1 Vaccination remains an important preventive measure against covid-19. Since the rollout of covid-19 vaccines in late 2020, global vaccine administration has accumulated up to 11 billion doses, with 13.08 million being administered daily.2 WHO has authorised the emergency use of 10 vaccines developed by Janssen, Bharat Biotech, Pfizer-BioNTech, Oxford-AstraZeneca, Moderna, Sinopharm, Sinovac, Novavax, and Serum Institute of India.3 Despite a rapid decline in the number of covid-19 symptomatic infections and deaths, several studies have raised concerns about waning vaccine induced immunity in vaccinated populations due to time and the emergence of covid-19 variants, which prompts the urgent need for a booster dose.45678 Furthermore, heterologous vaccine regimens could be an alternative strategy to homologous regimens when supplies are limited. Inconsistent covid-19 vaccine procurement and limited vaccine supply have resulted in certain vaccine types being unavailable in clinical settings.910 Research that evaluates different vaccine regimens will aid decision making in public health policy and reduce vaccination hesitancy.

On 16 December 2021, WHO released interim guidance summarising the existing evidence on heterologous covid-19 vaccine schedules and gave recommendations for these schedules.11 Although recent observational studies and systematic reviews have respectively assessed the clinical efficacy of covid-19 vaccines and provided a descriptive qualitative overview of various heterologous regimens, a quantitative comparison of different vaccine regimens against the original and new circulating variants of concern (such as omicron BA.1 and BA.2) is urgently needed.1213 Such a comparison is particularly important because randomised controlled trials that involve several vaccine types need to be multinational in scale and data are not readily available. Additionally, the effectiveness of a mixed combination of covid-19 vaccines remains uncertain. Quantitative systematic analyses of different vaccine regimens and a dynamic platform to monitor the effectiveness of various vaccine regimens in protecting against future variants of concern are urgently needed.

We compared the vaccine effectiveness of heterologous and homologous regimens with and without boosting in our living systematic review and network meta-analysis. Our study supplemented WHO’s summary report by quantitatively evaluating different covid-19 vaccine regimens: heterologous prime boost, single dose, homologous two dose, heterologous and homologous third dose boosting, with the no-vaccine group as a reference. The advantage of a network meta-analysis compared with a conventional meta-analysis is the high comparability of direct and indirect evidence, which enables vaccine effectiveness to be compared across pairs of studies, resulting in a more comprehensive interpretation of the available evidence. With network meta-analysis, we were able to summarise the effectiveness of all available covid-19 vaccine regimens and determine the relative effects of various primary and boosting regimens as assessed in current clinical trials.

Overall, our study will serve as a monitoring platform for informing the public and health officials about the vaccine effectiveness of all WHO recommended vaccines and their homologous and heterologous regimen combinations against circulating SAR-CoV-2 (current and future variants of concern). This study is ongoing and will be updated through this living systematic review.


This living systematic review and network meta-analysis followed the preferred reporting items for a systematic review and meta-analysis of network meta-analysis (PRISMA-NMA). Supplementary table 9 presents the PRISMA-NMA checklist.

Search strategy and selection criteria

We searched 38 WHO covid-19 databases for published studies and preprint databases on a weekly basis from 8 March 2022. No language restrictions were applied to the search. Supplementary table 1 gives the full search strategy. We followed prespecified inclusion criteria during study screening (supplementary protocol 2.1): studies that assessed the efficacy or effectiveness of covid-19 vaccines in humans; studies that investigated documented, symptomatic, severe covid-19 infections, covid-19 related hospital admissions, or covid-19 related deaths; commentaries, editorials, and correspondence were included if sufficient data were provided in a supplementary file. Populations of all ages and both sexes were included in this analysis. Age was stratified into three groups: young (<18 years), adult (18-64 years), and older (≥65 years). Exclusion criteria were applied in the network meta-analysis: one arm studies were excluded; studies that did not report vaccine efficacy or effectiveness were excluded.

WYA and PPHC independently performed a study search and screened titles and abstracts of all retrieved studies in EndNote 20. Retrieved studies were further assessed for eligibility using full text screening by the same reviewers. All disagreements were resolved by consensus between WYA and PPHC. Duplicated results were removed upon reference importation in EndNote 20 by WYA. Any remaining duplicates were eliminated manually.

Data synthesis

For every eligible study identified from full text screening, WYA and PPHC independently extracted information on the study characteristics: author and year, participant eligibility, age of participants, the proportion of male participants, distribution of baseline characteristics, vaccine priority groups, ethnicity, country of study, SARS-CoV-2 variants of concern investigated, an overall sample size of the study, trial registry for randomised controlled trials, study design, research aim, intervention group (treatment 1), comparator group (treatment 2), dose interval, follow-up period, clinical outcome assessed, and outcome measures (supplementary table 2). WYA and PPHC also extracted the respective number of events in the intervention and comparator groups and reported vaccine efficacy or effectiveness. When the number of events was not provided, the figure was derived using the reported odds ratio, risk ratio, incidence rate ratio, or hazard ratio, given that the total number of participants in each intervention and comparator group was known. For studies that recorded the number of events at two or more time points, data were extracted for the period when the vaccine was the most effective.

We estimated the overall effectiveness of each vaccine regimen (1−odds ratio). We created league tables that present relative effects in pairwise comparisons with 95% credible intervals. To rank the vaccination regimens with different combinations of vaccines, we determined the surface under the cumulative ranking curve (SUCRA) scores. To combine randomised and non-randomised evidence in the network meta-analysis, we estimated treatment effects using a three level bayesian hierarchical modelling approach with random effects (supplementary protocol 2.4.2).1415 We assumed that all studies shared a common heterogeneity variance. Vague priors were used for heterogeneity variance and treatment effect estimates. The number of iterations, burn-in, and adaptation used in the Markov chain Monte Carlo method is described in the protocol. JAGS was used to implement the bayesian hierarchical model.16 BUGS codes were provided on GitHub (

Network meta-analysis was performed twice, each with nodes defined in two different ways—vaccine product based and platform based—providing two perspectives on vaccine effectiveness. For the vaccine product based network, a node was made of vaccines of the same brand with the same number of doses. For the platform based network, vaccines of the same platform but different brands were grouped into the same nodes, given the number of doses was the same. Inconsistency in the networks was evaluated using the guideline developed by Daly and colleagues.17 Finally, we performed subgroup analyses by reanalysing studies that investigated the variable of interest (age, ethnicity, immunocompromised or not, or covid-19 variant) with all other factors controlled. Sensitivity analysis was done by restricting the analysis to low risk of bias studies.

For quality assessment of non-randomised trials, the risk of bias within individual studies was evaluated using the ROBINS-I tool (risk of bias in non-randomised studies of interventions), which was recommended by Cochrane reviews.18 The ROB-2 tool (Cochrane risk of bias version 2) was used to assess randomised controlled trials.19 We assessed the quality of the evidence by applying the GRADE method (grading of recommendations assessment, development, and evaluation) and gave a rating to each estimate obtained in our network meta-analysis.20 Publication bias in our analysis was assessed through a comparison adjusted funnel plot. Each data point in the funnel represented a pair of comparisons of treatments instead of a single study.21 The plot was drawn with the function netmeta::funnel.

Patient and public involvement

Many discussions with the public, such as the media, doctors, and patients, on their queries on the need for a booster vaccine dose have inspired this review. However, there is no direct patient and public involvement because our analysis does not require their involvement. We spoke to patients with covid-19 about the study, and we asked several public members to read our article after submission.


Study characteristics

Study selection followed PRISMA-NMA guidelines (fig 1). We identified 12 962 studies from 38 databases and removed 5559 duplicates, retaining 7403 studies for full text screening. We excluded 6923 studies by title, abstract, and subheading screening, of which 3007 (43.4%) were non-vaccine studies, and 1398 (20.2%) were studies of viruses other than SARS-CoV-2, 800 (11.6%) were reviews, and 502 (7.3%) investigated non-human subjects. The remaining 1216 studies (17.6%) were descriptive literature with no supplementary data. During the full text screening and data extraction, we excluded 404 studies. Twenty seven (6.7%) were protocols, seven (1.7%) were vaccine safety studies, and 370 (91.6%) only examined immunogenicity and reactogenicity in people who were vaccinated. Of the remaining 76 studies, we were able to extract data from 53,2223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374 which gave us a sample size of 100 190 476 participants from 15 countries. Supplementary table 2 presents a summary of the study characteristics. Eight studies included participants older than 652731434748556869 and seven studies enrolled participants younger than 18.45535760616873 Seven studies included high risk populations (defined as having more than one chronic condition),27404143474869 of which two studies included patients who were immunocompromised in a subset of the high risk populations.4869 Of the 53 included studies, 19 investigated the protectiveness of vaccines against covid-19 variants of concern or variants of interest, including alpha (B.1.1.7), beta (B.1.351), gamma (P.1), delta (B.1.617.2), and omicron (B.1.1.529).22263032343738394142444748495761636870

Fig 1
Fig 1

Flowchart of study selection

Risk of bias

We evaluated the risk of bias by following instructions in ROB-2 for randomised controlled trials and ROBINS-I for non-randomised studies.1819 Of the 37 non-randomised studies, 16 were rated to have a moderate risk of bias, mainly because they did not control for confounders such as comorbidities and other baseline characteristics.22323334353738444650566061656874 Nine studies from the same pool were also prone to high selection bias for having an imbalanced proportion of participants of different ages and sexes.223233343537505661 Three non-randomised studies relied on surveillance data, which were subject to incomplete information, and so received a moderate risk of bias score in the domain of bias due to missing data.384668 Finally, one study was rated as having a severe bias in selecting the reported result because the authors reported the overall vaccine effectiveness for mRNA vaccines instead of the specific vaccine products investigated in the study.44 All randomised controlled trials were considered low risk of bias except for three studies in which participants were unblinded after the second dose.404347 Supplementary figure 2 and supplementary table 4 show the results of ROB-2 and ROBINS-I.

Combinations of vaccine regimens in networks

There were two network analyses in this study: vaccine product based and platform based. In the vaccine product based network, we identified 24 covid-19 vaccine combinations from the 53 included studies and coded them by the number of doses used and the acronym of the vaccines (supplementary table 3). For example, 1AZ1BNT represented a heterologous prime boost regimen using ChAdOx1 (Oxford-AstraZeneca) as the first dose and BNT162b2 (Pfizer-BioNTech) as the second dose. In the platform based network, we identified 13 vaccine regimens where vaccine products of the same platform were grouped into the same node in the second network. Network diagrams for all five outcomes were drawn to depict the relation between all regimens. Most studies compared mRNA vaccines in the vaccine based and platform based networks, as indicated by the thickest lines (supplementary fig 1A-J). There was no disconnection between nodes.

Effect of vaccine regimens against documented covid-19 infections

Thirty five studies contributed to the investigation of vaccine effectiveness against documented covid-19 infections, of which 11 were randomised controlled trials. Supplementary tables 5 and 6 present the results of relative treatment effects for all pairwise comparisons of vaccine regimens. The odds ratios of all regimens analysed are smaller than 1 compared with the no-vaccine group, indicating that any vaccine regimen can confer protection against covid-19 infections, either asymptomatic or symptomatic. For three dose regimens in the vaccine product based network, two dose BNT162b2 with one mRNA-1273 (Moderna) booster is the most effective regimen (odds ratio 0.022, 95% credible interval 0 to 2.914), followed by three dose BNT162b2 (0.036, 0 to 3.033). However, the actual estimate for three dose homologous BNT162b2 is uncertain due to selection bias in one study that assessed BNT162b2 regimens.32 A two dose CoronaVac (Sinovac) with one dose BNT162b2 regimen is also highly effective (0.046, 0.004 to 0.567). Among the two dose regimens, heterologous primary vaccination with ChAdOx1 and BNT162b2 is more effective than two dose BNT162b2 or two dose mRNA-1273, with an average odds ratio difference of 0.04 (supplementary table 5A).

In the platform based network, three dose mRNA and two dose inactivated vaccine with one dose mRNA regimens are comparably effective (0.046, 0.009 to 0.275 for homologous mRNA; 0.047, 0.003 to 0.675 for heterologous inactivated). Two dose adenovirus with one dose mRNA is slightly less effective but prevents covid-19 better than any homologous two dose regimens (0.066, 0.007 to 0.541 for heterologous adenovirus; supplementary table 6A). One dose regimens (mRNA or adenovirus vector or protein based) are less protective than two dose regimens (mRNA or adenovirus vector or inactivated or protein based). Table 1 shows the SUCRA scores.

Table 1

Ranking of vaccine regimens

View this table:

Effect of vaccine regimens against symptomatic covid-19 infections

We were able to pool results from 23 studies that evaluated the vaccine effectiveness against symptomatic covid-19 infections, of which three studies assessed the effect of homologous boosters with BNT162b2, mRNA-1273, or ChAdOx1.252632 The odds ratios between a homologous booster dose and the no-vaccine group are 0.175 (95% credible interval 0.021 to 1.388), 0.104 (0.016 to 0.733), and 0.147 (0.017 to 1.240) for three dose mRNA-1273, three dose BNT162b2, and three dose ChAdOx1, respectively (supplementary table 5B). Three dose mRNA-1273 and three dose ChAdOx1 are more effective than the two dose homologous regimen of mRNA-1273 and ChAdOx1 with reference to the no-vaccine group. In the platform based network, three dose mRNA is shown to be the most effective among all regimens (0.019, 0.002 to 0.136; supplementary table 6B). The one dose protein based regimen appeared to be highly effective (0.094, 0.001 to 10.134). Still, the studies that investigated this regimen were found to be confounded by unblinding and unexplained censoring of data.4047 Therefore, certainty in the estimate for a one dose protein based vaccine regimen is low.

Effect of vaccine regimens against severe covid-19 infections

We analysed 12 studies for vaccine effectiveness against severe covid-19 infections. With reference to the no-vaccine group, all one dose regimens are less effective than two dose and three dose regimens, of which one dose Ad26.cov2.S (Janssen) and one dose CoronaVac had no association with a reduction of severe covid-19 infections (odds ratio>1; supplementary table 5C). Three dose BNT162b2 was more effective compared with any vaccine regimens (odds ratio<1 in all comparisons; supplementary table 5C).

Effect of vaccine regimens against covid-19 related hospital admission

Seventeen studies were evaluated for vaccine effectiveness against covid-19 related hospital admissions. Individuals receiving three dose BNT162b2 or three dose mRNA-1273 are the least likely to be admitted to hospital because of covid-19 (odds ratio 0.054, 95% credible interval 0.010 to 0.294 for three dose BNT162b2; and 0.018, 0.001 to 0.291 for three dose mRNA-1273; supplementary table 5D). Results showed that heterologous primary vaccination of ChAdOx1 with BNT162b2 is not as effective as two dose homologous BNT162b2, mRNA-1273, or ChAdOx1 regimens (0.515, 0.091 to 3.291 for heterologous primary ChAdOx1 with BNT162b2; supplementary table 5D). However, the estimates for two dose BNT162b2 and mRNA-1273 could be biased by confounders in the studies that assessed them; therefore, the true estimates of odds ratios for the two dose homologous mRNA vaccine regimens might be higher. Studies that reported covid-19 hospital admissions are mainly observational, which also adds uncertainty to the estimates.

Effect of vaccine regimens against covid-19 related deaths

Eight studies were evaluated for vaccine effectiveness against covid-19 related deaths. Estimates for the prevention of deaths are highly uncertain because observational studies were the only evidence available in this analysis. None of the randomised controlled trials reported deaths. Results could be confounded by age and disease conditions, leading to highly uncertain estimates.

Subgroup analyses

Owing to high uncertainty in the estimates of vaccine effectiveness and limited data availability for severe covid-19, hospital admissions, and deaths, we only performed subgroup analyses on studies that investigated non-severe SARS-CoV-2 infections, asymptomatic and symptomatic. We were able to stratify studies by age, being immunocompromised or not, and covid-19 variants. Sex and ethnicity were not investigated because of limited data.

We found that three dose regimens conferred protection to all age groups (young, adult, and older). However, the young group appeared to have slightly better immunity than the adult group (odds ratio difference 0.02) and the older group (odds ratio difference 0.07) after receiving a three dose regimen (supplementary table 8A). On average, the vaccine effectiveness of any regimen was lowest in the older group. A two dose CoronaVac regimen was found to have no association with covid-19 infections in the older group (odds ratio 1.13, 95% credible interval 0.22 to 16.92; supplementary table 8A). Supplementary table 8B shows comparisons of vaccine effectiveness in patients who are immunocompromised. A three dose mRNA regimen worked comparably well in the non-immunocompromised and immunocompromised groups. However, the effect of two dose mRNA vaccines is weaker in those who are immunocompromised than in those who are non-immunocompromised (supplementary table 8B).

For the covid-19 variant subgroups, sufficient data were only available to analyse alpha, delta, and omicron variants. Vaccine regimens with two or more doses were found to be effective against the alpha variant. The effectiveness of three dose vaccine regimens within variants is comparable; however, effectiveness between variants differs considerably. The three dose regimens are less effective in preventing delta and omicron infections (supplementary table 8C). All one dose and two dose regimens appear to be ineffective against omicron. The estimates in the variant subgroups could be underestimated owing to a surge in confirmed infections during the outbreak of omicron; therefore, we have low certainty on the true vaccine effectiveness against this variant. We reanalysed the vaccine effectiveness in a sensitivity analysis without all high risk of bias studies. Table 2 shows the results.

Table 2

Odds ratios (95% credible intervals) and vaccine effectiveness of vaccine regimens by platform with only low risk of bias studies

View this table:

Inconsistency assessment of network

We assessed inconsistency in the vaccine product based and platform based networks by comparing residual deviance between the inconsistency and consistency model. The deviance contribution plot shows some points below the line of equality (supplementary fig 3A, B). Further assessment of inconsistency was done using the node splitting model (supplementary table 7).

Publication bias

Publication bias was examined by using the comparison adjusted funnel plot. We hypothesised that published studies tend to report better results than unpublished studies. A comparison adjusted funnel plot coupled with Egger’s test was used to detect a small study effect (supplementary fig 4). Although a few estimates were lying away from the centre, Egger’s test P value indicated that statistical significance was not reached (P=0.08).


Since the launch of covid-19 vaccines in 2020, research efforts have been made to investigate different combinations of covid-19 vaccines as alternatives to homologous regimens. This review has provided a comprehensive analysis of the effectiveness of WHO approved vaccines and compared all available vaccine regimens. We assessed vaccines of different brands and platforms. Comparisons by platforms are more informative and translatable into practice because vaccines from different manufacturers have been shown to have similar efficacy in phase trials.274041454755637173 Our findings will serve as a reference for clinicians, public health policy makers, and researchers for vaccine related purposes, such as making recommendations to patients and public health decision making.

Principal findings

We compared the vaccine effectiveness in preventing five outcomes: covid-19 related documented infections, symptomatic infections, severe infections, and covid-19 related hospital admissions and deaths. We have higher certainty in the evidence relating to covid-19 infections than for covid-19 related hospital admissions and deaths owing to a high risk of bias and imprecision. We have low certainty in the true estimates for vaccine effectiveness against the omicron variant because the relevant studies were conducted during the omicron outbreak. However, we are confident that a three dose regimen will effectively prevent covid-19 variants. The results have consistently shown a considerable reduction in covid-19 infections across different variants, despite the studies conducted during the peak of infections.

Three dose mRNA vaccines (three dose BNT162b2 or mRNA-1273) appear to be the most effective in preventing non-severe covid-19 infections. A heterologous regimen with an mRNA booster in recipients of two dose adenovirus vector vaccines also has 80% protection against covid-19. Among all two dose regimens, mRNA vaccines remain the gold standard for prevention against all covid-19 related outcomes, although we have moderate certainty in the actual effectiveness against hospital admission caused by covid-19. Our results imply that mRNA vaccines will continue to be the preferred vaccine type, either as primary vaccines or booster doses. When comparing vaccine effectiveness between age groups, we found people younger than 18 have a lower chance of covid-19 infection after receiving vaccines of any platform. This finding agrees with a recent immunogenicity study in children and adolescents.75 We also found that a heterologous or a homologous third dose booster can confer an equal level of protection in all age groups, even in the older group (≥65). If several boosters are administered to any age group, a heterologous or homologous regimen will not make much difference in immunity improvement. Importantly, we discovered that any homologous or heterologous three dose regimen induces a considerably higher level of immunity in older people than a two dose homologous regimen, which is comparable to that seen in all other age groups.

Vaccine effectiveness in patients who are immunocompromised

In the vaccine effectiveness comparison between immunocompromised and non-immunocompromised groups, we found that all vaccine regimens have lower effectiveness in people who are immunosuppressed or immunodeficient, with a decrease of 2% vaccine effectiveness. Similarly, a review paper has shown weaker vaccine induced immunity among people who are immunocompromised, such as lower seroconversion rates.76 Our review suggests that a third booster dose, as part of a heterologous or homologous regimen, will greatly improve protection in these patients compared with a two dose primary vaccination. This finding orthogonally agrees with a randomised trial that studied SARS-CoV-2 antibody seroconversion in people who were immunocompromised and received a heterologous or homologous booster after primary mRNA vaccination. Therefore, the number of doses of vaccines seems to be the key to improving immunity rather than the combinations of vaccine types.77

Vaccine effectiveness against covid-19 variants

Rapidly evolving viral strains continually pose challenges to the elimination of covid-19. Recent immunogenicity research has reported waning vaccine effectiveness against delta and omicron variants.78 Our estimates showed that the two dose and one dose regimens were ineffective against the omicron variant. These estimates are in line with the findings on omicron variant neutralisation after a homologous two dose mRNA regimen.79 However, our study shows that homologous and heterologous three dose regimens successfully reduce covid-19 infections caused by the omicron variant. One study has shown that people receiving an mRNA booster after two doses of CoronaVac have a 1.4-fold increase in neutralisation activity against omicron.80 Therefore, boosting vaccination will effectively control the spread of covid-19 variants. The latest study of the fourth dose of BNT162b2 reported that effectiveness against confirmed infection and severe covid-19 improves from receiving a third dose to a fourth dose in people aged 60 or older.81 This finding implies that ongoing vaccine campaigns will be needed to prevent covid-19 infections in the long term. According to our results, mRNA vaccines appear to be the preferred choice for any additional dose.

Although we could not pool the results from 23 studies, all studies suggested that people who received a third dose mRNA or heterologous boosting regimen are less likely to get infected by SARS-CoV-2 than those receiving only a primary homologous regimen. When considering the safety of heterologous and booster vaccines, one study that assessed the safety and reactogenicity of heterologous primary vaccination with mRNA, adenovirus vector, and protein based vaccines showed no safety concerns.82 Another study that examined the safety of booster doses in adults also showed fewer local and systemic reactions after a homologous mRNA booster than after a two dose homologous regimen.83

Limitations of study

Our study did not evaluate the optimum time interval for prime boost or boosting regimens owing to limited information about the dynamics of vaccine effectiveness across a period in a few studies. However, we anticipate more longitudinal research on the varying infection rates among people who are vaccinated and those who are not vaccinated. This type of study that combines timely measurement of antibody titres will provide more evidence on how the impact of vaccination changes over time and the protection period of a series of vaccinations.


A three dose mRNA regimen seems to be the most effective in preventing covid-19 infections. An mRNA booster can induce a similar level of protection against covid-19 infections to homologous primary vaccination. A third dose vaccine is needed to prevent covid-19 variant infections. Heterologous and homologous three dose regimens work equally well in preventing any covid-19 infections, even variants. We will update the results when newly published studies or preprints become available. For example, we will add other vaccine types and multiple dose regimens to the analysis as more vaccines are approved by the WHO emergency use listing. More research on multiple doses of the primary vaccination is expected. We will also examine the efficacies of vaccine regimens against new variants for the general population and other subgroups, such as sex, ethnicity, and other high risk populations.

What is already known on this topic

  • The efficacy and effectiveness of individual vaccine products for covid-19 and its variants of concern are well known

  • Research on the effectiveness of vaccine combinations, especially for particular populations such as older people and those who are immunocompromised, is lacking

What this study adds

  • This living systematic review and network meta-analysis investigated the effectiveness of different homologous and heterologous vaccine regimens with and without boosting against covid-19 infections and covid-19 related hospital admissions and deaths

  • An mRNA booster to any primary vaccination included in our study confers a high level of protection similar to a homologous three dose mRNA regimen; a third dose is needed to prevent infection caused by the omicron variant

  • Any homologous or heterologous three dose regimen induces higher immunity in people ≥65 than a two dose homologous regimen; a three dose mRNA regimen reduces the risk of asymptomatic or symptomatic covid-19 infections in the immunocompromised population

Ethics statements

Ethical approval

Not required.

Data availability statement

Raw data in this systematic review with meta-analysis are extracted from published and preprint studies available on the internet. Our processed data for network meta-analysis and R codes on GitHub (


We thank Nick Chow, Bryan Fu, Jet Leung, and Charmine Tsang for their assistance with the study search and data extraction.


  • Contributors: PPHC conceived and supervised the study. WYA and PPHC initiated and designed the study plan. WYA and PPHC performed the search, independently screened search results and extracted data from eligible studies. WYA and PPHC exchanged data extraction results and resolved disagreement through discussion. WYA performed the statistical analysis. WYA and PPHC wrote the manuscript. We confirm that both authors had full access to all data in the study and final responsibility for publishing it. WYA and PPHC accessed and verified the data. WYA and PPHC are study guarantors. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

  • Funding: This work was supported by the Chinese University of Hong Kong Department of Chemical Pathology’s Faculty Startup Fund. The funder had no role in study design or the collection, analysis, interpretation of data, writing of the manuscript, or decision to submit the article for publication.

  • Competing interests: All authors have completed the ICMJE uniform disclosure form at and declare: support from the Chinese University of Hong Kong for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous three years; no other relationships or activities that could appear to have influenced the submitted work.

  • The lead authors affirm that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

  • Dissemination to participants and related patient and public communities: We will disseminate the results to clinicians, patients, governmental organisations, and agencies through social media and press releases.

  • Provenance and peer review: Not commissioned; externally peer reviewed.

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: