A review on COVID-19 transmission, epidemiological features, prevention and vaccination

Airborne transmission

The COVID-19 transmission has been confirmed to occur during close person-to-person contact, via respiratory droplets (>5–10 µm in diameter) and fomites from coughing or sneezing, similar to the transmission modes of SARS and MERS [38]. A study recently performed by Lelieveld et al. [39] estimated the infection risk for different indoor environments such as an office, a classroom, choir practice, and a reception/party. It showed that aerosols from super infective subjects (the top 5–10% of subjects with a positive test) can effectively cause COVID-19 clusters (>10 infections) in indoor environments [39].

As a direct dispersion of the SARS-CoV-2, microscopic aerosol particles (<5–10 µm in diameter) consisting of the evaporated respiratory droplets can remain airborne for 3 h, and the airborne transportation distance can reach 7–8 m [40, 41]. Si et al. investigated the behavior and fate of respiratory droplets (0.1–4 µm in diameter) during coughs in a single-path respiratory tract model. 2 µm droplets have the highest exhalation fraction from the pulmonary alveoli. The highest velocity of the exhaled droplets can reach 50 m/s for individual droplets [42]. Microscopic aerosol particles can be inhaled and penetrate deeply into the alveolar region of the lungs, readily causing COVID-19 infections in the indoor environment [39].

Alternatively, SARS-CoV-2 may combine with ambient aerosols and enter the human body. There are two possible mechanisms of ambient aerosols on the transmission of COVID-19 [41]. Firstly, ambient aerosols can be the carrier of COVID-19 to enter the human body. Secondly, the ambient aerosol can stimulate the expression of the SARS-CoV-2 receptor to accelerate the infection efficiency [41].

Faecal-oral transmission

Multiple studies highlight the threat of potential faecal-oral viral transmission of COVID-19 [43, 44]. In a recent study by Zuo et al. [43], faecal samples of 15 hospitalized patients with COVID-19 were collected, and it was found that 7 of 15 patients had stool positivity for SARS-CoV-2 by RNA shotgun metagenomics sequencing. The faecal viral metagenome of three patients continued to display active viral infection signature up to 6 days after clearance of SARS-CoV-2 from respiratory samples [43]. In addition, faecal samples with a signature of high SARSCoV-2 infectivity harbored a higher abundance of opportunistic pathogens, but live viruses were not to be able to be titrated or isolated from the faeces due to the methodological limitation [43]. A retrospective cohort study collected 3,497 respiratory, stool, serum, and urine samples from 19 January 2020 to 20 March 2020, to evaluate viral loads at different stages of disease progression in patients infected with SARS-CoV-2 in Zhejiang province, China [44]. RNA was detected in the stool of 55 (59%) patients and the serum of 39 (41%) patients. The urine sample from one patient was positive for SARS-CoV-2 [44]. Xing et al. [45] collected three pediatric cases of COVID-19 reported from January 17, 2020 to February 23, 2020 for Real-time fluorescence reverse-transcriptase-polymerase-chain reaction (RT-PCR) to detect SARS-CoV-2 RNA in throat swabs and fecal specimens. Viral RNA remained positive in the stools of pediatric patients after clearance of SARS-CoV-2 in the respiratory tract [45]. Two children had fecal SARS-CoV-2 turned negative 20 days after throat swabs showing negative, while that of another child lagged for 8 days [45]. These results suggested the potential for the virus to be transmitted through fecal excretion as well. As some studies reported to have prolonged viral RNA in rectal faecal samples [45, 46] more evidence is clearly required to accurately reveal the infectivity and pathogenesis of faecal-oral transmission of SARS-CoV-2 [43, 47].

Vertical transmission

Vertical transmission is defined as the transmission of the infectious pathogen from the mother to the fetus during the antepartum and intrapartum periods, or to the neonate during the postpartum period via the placenta in utero, body fluid contact during childbirth, or through direct contact owing to breastfeeding after birth [48]. There is limited but increasing evidence of SARS-CoV-2 infection during pregnancy. Several papers report the detection of SARS-CoV-2 RNA in pregnant women who received a coronavirus disease 2019 diagnosis [49], [50], [51]. A systematic review and meta-analysis of such studies found that a pooled proportion of 3.2% of neonates from mothers with COVID-19 have a positive result for severe acute respiratory syndrome coronavirus 2 viral RNA test using a nasopharyngeal swab. Severe acute respiratory syndrome coronavirus 2 viral RNA testing was positive in 2.9% of neonatal cord blood samples, and 7.7% of placenta samples [48]. Another systematic review on the possible transmission of SARS-CoV-2 through breast milk and breastfeeding found that break milk samples from 44.0% of the mothers diagnosed with COVID-19 were reported to be positive for SARS-CoV-2 RNA via RT-PCR analysis [52]. But among the cases with viral RNA detected in breast milk samples, one healthy neonate was fed with a breast milk substitute, while the maternal breast milk samples had detectable viral RNA and thus, it is not possible to ascertain the risk of infection by exposure to breast milk [51]. Given that most neonates born to mothers diagnosed with COVID-19 by RT-PCR tests during their pregnancy were negative for viral infection, and the majority of included studies did not assess the presence of SARS-CoV-2 specific antibodies in neonates, breast milk, placenta, or other tissues and hence, the evidence of possible SARS-CoV-2 transmission through breast milk is still limited [52]. Studies are needed with longer follow-up periods that collect data on infant feeding practices and viral presence in breast milk.

Search strategy and selection criteria

This systematic review and meta-analysis were conducted to estimate the pooled R₀ of COVID-19 in different counties via articles published in international journals. We searched PubMed, Web of Science, bioRxiv, and arXiv to obtain studies published between 1 December 2019 and 21 March 2021 regarding the reproduction number of COVID-19. We used the search terms or combinations of terms as the following – a) “Coronavirus Disease 2019” OR “COVID-19” OR “SARS-CoV-2”, b) “model” OR “modeling”, c) “reproductive number”. Studies were included if they presented a mathematical/statistical model of SARS-CoV-2 and reported R₀ and at least one of the following parameters—R_t, incubation period, generation time, latent period, or serial interval. Studies were excluded if they met the following conditions: a) Studies had conceptual confusion surrounding R₀ and R_t, b) Studies did not report necessary parameters like R₀ with its 95% confidence interval (CI), c) Studies with R_t did not definitely discuss the interventions or the implementing period of the intervention was not reported, d) Simulation studies on R₀/R_t, e) Reviews comments, reports, and letters, f) Articles not in English. Three reviewers (the second author CSR, the third author JX and the fourth author WY) independently performed the literature search and screened titles and abstracts to exclude studies that did not meet our selection criteria before the full review of the full text. Disagreements were resolved by discussion among all authors.

Data extraction and statistical analysis

We used a predefined standardized extraction table to extract data. The extracted variables included the name of the first author, area of the study, period of data, model type and settings, estimated R₀ (with a 95% confidence interval, CI), estimated R_t/R_e/R_c (with a 95% confidence interval, CI), and public health measures in countries if available. Data extraction was performed by CSR, JX and WY independently, and results were summarized by the co-first author ZYQ and WGH. Synthesis was carried out to aggregate the information from multiple studies with the same estimates (or effect size of interest) yet different features including the differences in the data collection, the sample size, and the conditions. A meta-analysis was conducted to obtain overall R₀ and continental R₀ values.

Heterogeneity between studies was assessed using the Cochran Q test, the I² index, and T² index (Table 1). According to the I² results, heterogeneity can classified into the following three categories: I²<25% (low heterogeneity), I²=25–75% (average heterogeneity), and I²>75% (high heterogeneity) [71]. Because of the high I² values that were calculated (99.1% for the globe, 99.4% for Asia, 98.4% for Europe, 90.9% for Africa, 97.6% for South America, and 84.1% for North America), as well as the significance of the Cochran Q test (p<0.0001), a random-effects model was used to estimate R₀ in this work. Publication bias was assessed using Begg and Egger tests, as well as the funnel plots [72]. The bias was adjusted with trim and fill method [73]. Moreover, leave-one-out sensitivity analysis was also employed to assess the robustness of the results. Data were analyzed using R version 3.6.2 [74].

Results of COVID-19 R₀ estimates

We identified 1,350 studies, of which 71 were duplicates, which left us with 1,279 reports. A total of 1,039 reports passed the initial title screening, and 718 records were excluded by abstract, leaving 321 unique studies. Of these, 291 were excluded because they were simulation studies or did not report necessary parameters like R₀ with its 95% confidence interval (CI). 30 publications passed the full-text assessment for eligibility (Figure 1). The 30 included studies with a total of 68 R₀ records used a broad range of methods to generate R₀. 63.3% (19/30) were compartment models, among which 26.3% (5/19) and 57.9% (11/19) of models belong to Susceptible-Infected-Recovered (SIR) models and Susceptible-Exposed-Infectious-Recovered (SEIR) models, respectively. 36.7% (11/30) of reviewed papers used other kinds of models such as the networked dynamic metapopulation model [75], the generalized growth model (GGM) [76, 77], the extended and modified Flaxman model [78], etc. 30 studies included in the meta-analysis reported estimated R₀ and its 95% CI (Table 2). All the reports were conducted in 43 countries and regions in 2020, including much of East Asia, South Asia, and Europe, and a handful of countries in Africa and North and South America. The proportion of reviewed records in these five continents were shown in Figure 2.

Figure 1:

Flowchart of reports selection for inclusion in the systematic review.

Figure 2:

The proportion of records reporting R₀ values in five continents. R₀, the basic reproduction number.

The median R₀ was 2.9 (IQR, 2.1–3.8) for all the reviewed studies. After stratified by continents, median R₀ for Asia, Europe, Africa, South America, and North America over the whole outbreak periods were 2.4 (IQR, 2.1–2.6), 3.3 (IQR, 2.4–4.3), 3.2 (IQR, 2.8–3.5), 2.6 (IQR, 2.4–3.1) and 2.4 (IQR, 2.1–2.6), respectively. For publication bias assessment, the funnel plot (see Figure S1, Supplementary material), Begg test (p<0.0001), and Egger test (p=0.0241) suggested significant bias. The sensitivity analysis showed that no individual study significantly affected the pooled results of R₀ (see Table S1, Supplementary material, which illustrates the sensitivity analysis results of meta-analysis for R₀ estimates). According to the results of the random-effects model, the pooled R₀ for COVID-19 was estimated as 2.9 (95% CI: 2.7–3.1), which means that each person infected with COVID-19 transmitted the infection to between 2 and three susceptible people on average. After stratified by continents, pooled R₀ estimates for Asia, Europe, Africa, South America and North America were 2.4 (95% CI: 2.1–2.7), 3.2 (95% CI: 2.9–3.5), 3.1 (95% CI: 2.3–3.9), 2.9 (95% CI: 1.8–4.0) and 2.4 (95% CI: 1.4–3.4), respectively (Table 1). We stratified the analysis by sources of cases to make the R₀ values comparable as well. Confirmed cases are people that have been tested, and the test confirms they have COVID-19 (i.e., a positive test). While the reported case figures on a given date do not necessarily show the number of new cases on that day – this is due to delays in reporting. The actual number of cases is likely to be much higher than the number of confirmed cases – this is due to limited testing. Confirmed cases have been used for R₀ estimates 20 out of the 30 studies (Table 2). As demonstrated in Figure 3A and B, the pooled R₀ estimate for confirmed cases (R₀, 2.9; 95% CI: 2.7–3.1) was higher than that for reported cases (R₀, 2.6; 95% CI: 2.3–2.9).

Figure 3a:

Meta-analysis of the synthetic estimated R₀ for COVID-19 with confirmed cases. R₀, the basic reproduction number. CI, confidence interval.

Figure 3b:

Meta-analysis of the synthetic estimated R₀ for COVID-19 with reported case data. R₀, the basic reproduction number. CI, confidence interval.

It is necessary to estimate the R₀ of COVID-19 to determine the capability of transmission per primary infected person to the secondarily infected persons [105], as well as to provide a reference for the government to make interventions to control the spread of the disease [106]. Our systematic review and meta-analysis found that the overall R₀ was 2.9 (95% CI: 2.7–3.1), which is similar to the results of an earlier review of 30 articles that were based on global level evidence (2.7) [10]. However, this estimate is lower than the previously summarized R₀ values for the COVID-19 (3.15 [13] and 3.32 [9]). Our estimation is similar to the R₀ values estimated for the MERS (R₀<1–3) [107, 108], and is higher than R₀ estimates for SARS (R₀=0.58–1.17) [109]. Since the number of daily reported cases does not necessarily reflect the number of new cases on that specific date, which is because of the long reporting chain that exists between a new case and its inclusion in national or international statistics, R₀ estimates were stratified by sources of cases. Our results showed that using reported cases in the early stage of epidemic can underestimate the R₀. In our subgroup synthetic analysis, R₀ values vary among different continents, that is R₀ estimates for Asia, Europe, Africa, South America, and North America varied from 2.4 to 3.1. It is noted that the total value of R₀ in a large population is the average of the R₀ subtypes, indicating R₀ is not an intrinsic characteristic of a given pathogen but rather describes the transmissibility of that pathogen within a specific population and setting [9]. In the most basic formulations of R₀, its value fundamentally depends on three primary parameters: the duration of contagiousness from the infection onset, the probability of infection for each contact between a susceptible person and an infected person, and the contact rate [110]. In other words, R₀ can be viewed as a combination of the pathogen’s biological constants and factors potentially influencing the contact rate, such as population density, social organization, and other epidemiological triads [61]. Because of the variation of social and human behavior among different areas, the transmission rate (β) and the recovery rate (γ) in epidemiological models are typically various. For example, Al-Raeei used eight countries’ reported COVID-19 data to fit the same SIRD (susceptible-infectious-recovered-death) model and found the coefficient of recovery range from 0.0008 in China to 0.0981 in India [111]. In addition to the socio-behavioral factors making the R₀ different in areas, the regional discrepancy of COVID-19 mutations may also produce different R₀. Mercatelli et al. suggested that a significant variation existed in the distributions of COVID-19 clades in different areas around the world [25]. Viruses mutations are of considerable medical and biological relevance as it has a great impact not only in the model estimation but also in the prevention and diagnosis of infectious disease. For instance, the early analyses announced by UK Prime Minister showed that a new strain of COVID-19 could increase the reproduction number by 0.4 or more [112]. Recently, the G614 mutation in the spike protein gene (D614G) has been calling scientific attention [113]. It was recently reported that this variant exhibits more efficient infection, replication, and competitive fitness in human airway epithelial cells, but maintains similar morphology and in vitro neutralization properties, compared with the ancestral wild-type SARS-CoV-2 virus [114]. A recent study found the potential link between viral genomic variation and its impact on transmission. The G614 mutation increases the rate of transmission of SARS-CoV-2, and hence the R₀ [115]. Thus, the mutation distribution of COVID-19 in different areas should also be accounted for in the variation of estimated R₀ if there are changes in the transmissibility of strains.

R₀ estimates may be somewhat error-prone for reasons such as data insufficiency and the short period analyzed in the early stage of COVID-19. As more studies are done and more data are produced, the hope is that this error will be reduced. Thus modeling structures and assumptions may be another reason for the discrepancy of R₀ across continents. The R₀ values can be estimated through a variety of models, the three most popular methods were the Stochastic dynamic model-based method, Poisson likelihood-based (ML) method, and Exponential growth rate-based (EGR) method [116, 117]. Very few confirmed cases were reported in the early stage of the COVID-19 epidemic, which made high-quality data rarely available for mathematical models. The accuracy of R₀ values mainly depended on model structures and assumptions about demographic dynamics. For instance, Exponential Growth (EG) and maximum likelihood estimation (ML) were applied to estimate the R₀ of 2019-nCoV using data of confirmed 2019-nCoV cases before January 23, 2020. The R₀ values estimated through the two different models were 2.90 (95% CI: 2.32–3.63) and 2.92 (95% CI: 2.28–3.67), respectively [118]. Hao showed the R₀ value was about 3.5 according to the MSEIR model and about 1.5 according to the MSIR model utilizing the COVID-19 data from January 21 to February 17, 2020 [119].

In addition, the estimation accuracy of R₀ was sensitive to the epidemiological parameterizations of models, which can be a reason for different R₀ estimates in the same countries as well. Mohammad et al. [8] demonstrated that COVID-19 R₀ is sensitive to the applied mean generation time when they estimated R₀ in Qom, Iran. Zhou et al. [105] showed the value of R₀ for SARS-CoV-2 was most sensitive to the generation time, so the more accurate estimation of the generation time based on the accumulation of epidemiological survey data would further improve the quality of the estimation of R₀. For other pathogens, R₀ values can be sensitive to model parameters as well. It was shown that R₀ for SARS was most sensitive to the transmission rate and the relative infectiousness after isolation in hospitals [109]. A study of Echinococcosis in Xinjiang, China found that the initial value S_H(0) can influence the R₀ most compared to S_D (0), I_D (0), and other initial conditions [120].

Comprehensive interventions

Some countries took comprehensive control interventions against COVID-19. A study on the impact of interventions in limiting COVID-19 transmission in Sichuan, China estimated the number of cases averted by the implemented control strategies [86]. The outbreak resulted in 539 confirmed cases, lasted less than two months, and no further local transmission was detected after February 27. It may be attributed to a strict set of measures by the Sichuan government to deal with the outbreak, including case isolation, tracing and screening of contacts of confirmed cases, quarantine of travelers from affected areas, and screening of people’s temperature in public places [86]. According to a study by Lemaitre et al., Large-scale non-pharmaceutical interventions (NPIs) were implemented by the cantons and the federal government between 28 February and 20 March 2020 in Switzerland, including a set of measures: (1) ban on gatherings of more than 1,000 people, (2) school closure, (3) closure of non-essential activities, and (4) ban on gatherings of more than five people [96]. As a result, the national R₀ was to be 2.8 (95% CI: 2.1–3.8) at the beginning of the epidemic. Starting from around 7 March, a strong reduction in reproductive number was found, with an 85.7% median decrease [96] (Table 3). Research suggests that the reproductive number of SARS-CoV-2 decreased by 95.6% since the adoption of NPIs, showing that NPIs were effective also in Slovenia for controlling SARS-CoV-2 (Table 3). R_t estimates of China, Japan, and 11 European countries were all below 0.5 (Figure 5), indicating that each sick person may infect fewer than 0.5 people on average, and the number of infected individuals may shrink over time. Generally, it will benefit a lot to implement early and continuing comprehensive interventions [123]. Nowadays, a large number of countries worldwide have conducted contact tracing to help identify infected people and contain the spread of COVID-19. In some countries, tracing “contact of a contact” has been used as well, which means any individual who is a contact of proximate contacts (i.e. spouse, children, co-workers, etc.) would be at risk and subject to quarantine once the proximate contacts had or developed symptoms, or tested positive for the virus causing COVID-19. This measure was always combined with modern technologies, such as cell phones, closed-circuit television (CCTV), and bank transactions, which helps collect information in an effective way, but privacy concerns and some ethical questions about exposing peoples’ activities and movements have been also raised [124].

Lockdown

Strict strategies of lockdown were imposed in some countries [80, 81, 99, 103]. For instance, a report of Chile on the effectiveness of lockdowns estimated that the reproductive number declined by 66.3% and reached 0.9 during lockdown (Table 3), indicating that the control measures at the start of the epidemic significantly slowed down the spread of the virus [103]. Large changes in R_t in response to the combined lockdown interventions were estimated for Tunisia as well [81]. The R_t moves to 0.89 (95% CI: 0.84–0.94) by the national lockdown measure, which is under the epidemic threshold 1 (Table 3). A nationwide lockdown restricting all nonessential movement throughout Greece began on March 23. A survey showed that before measures were implemented, the estimated R₀ was 2.38 (95% CI 2.01–2.80) [99]. During the lockdown, daily contacts decreased by 86.9% and reproductive numbers decreased by 81.0% (Table 3). A total lockdown intervention applied in Algeria was effective in Algeria with a reduction of transmission by 82.8% [80] (Table 3). As shown in Figure 5, R_t estimates of Algeria, Greece, Tunisia, and Chile taking a control measure of lockdown were all below 1. Practically, a lockdown decision can have very extensive effects, it is also dependent on how prepared a country is in the early phases of the pandemic [125]. Moreover, the uneven enforcement of central and local authorities in each country has led to different outcomes. Simply lockdowns without adequate testing, targeted quarantines or other interventions will not be effective in the long run [126].

Isolation and social distance

Many countries adopted control measures of isolation and social distance. A study in Santa Clara County, California, the US estimated a shift to partial social distancing, combined with rigorous testing and isolation of symptomatic individuals, is a viable alternative to indefinitely maintaining shelter-in-place. It also estimated that if Santa Clara County had waited one week longer before issuing shelter-in-place orders, 95 additional people would have died by April 22 [100]. Similarly, a report in Iran estimated the reproduction number as 4.4 (95% CI: 3.9–4.9) by using a generalized growth model, and the reproduction number was 1.55 after social distancing interventions were implemented, with a decrease of 64.8% [76]. Literature shows that if social distancing measures are not respected, a difference of at least 100,000 cases may occur over the next 300 days in Brazil [102]. A report in Jodhpur, Rajasthan, in the northwestern part of India suggests that control measures of isolation can reduce transmission by 29.0% [88], which is lower than that of the USA, Iran, and Brazil (Table 3). It is noted that implementing period of home isolation in India was 57 days, almost two times the period of quarantine in Brazil, but the reduction of reproduction number in India was only half that of Brazil (Table 3). It may owe to the poor consciousness among the people. They may not take it seriously and go back to taking part in social functions without adhering to protocols and appropriate behavior. Strengthening propaganda and following the COVID-19 appropriate behavior continuously, the effectiveness of control measures will get better. In summary, compared with comprehensive interventions and lockdowns, taking only a measure of isolation likely is much less effective for SARS-CoV-2, which agrees with studies by Davies et al. [127] and Flaxman et al. [98]. It is noted that R_t estimates of Iran, Brazil, India and USA taking control measures of isolation and social distance were around 1 or over.

Other public health interventions

Other protection measures were taken as well. The Japanese government declared a state of emergency on April 7, 2020 since the serious growth of the daily number of newly reported cases started in late March [128]. The R_t during the period of the state of emergency was estimated to be 0.36, with a reduction in the reproductive number of 86.2% (Table 3), suggesting the state of emergency might have been highly effective on the first wave of COVID-19 in Japan [90]. A study in Nigeria implied that the measure of international travel ban put in place have not been effective enough, because it could only help reduce the relative infectiousness of imported cases, with a 70.2% decrease of transmission (Table 3). Efforts must be concentrated on ensuring that the existing measures are improved and additional measures better than the existing ones can be introduced [79].

Live attenuated virus vaccine

This is the most traditional technology used for vaccine development. It can provide the most immunogenic vaccine formations since the weakened microbes retain the ability to replicate in vivo. It has the ability to stimulate the immune system by inducing the toll-like receptors (TLRs) that involve B cells, CD4, and CD8 T cells [132]. However, this comes at the cost of safety issues, including the possibility of reversion to a virulent state and the risk of significant side effects [133]. Thereby, this kind of vaccine needs to be carefully assessed before proceeding to clinical use, and their storage and handling require strict procedures.

The WHO listed three preclinical ongoing studies focusing on live attenuated vaccine development [134]. The Serum Inst of India in collaboration with Codagenix Inc is developing a live attenuated vaccine based on their CodaVax technology. Accessible safety and immunogenicity of the Influenza vaccine in animal models were documented by using this technology [135]. The University of Hong Kong also starts from an influenza-based vaccine strain with a deletion in the NS1 gene (DelNS1-SARS-CoV2-RBD). They re-organized the gene to express the RBD domain of SARS-CoV-2 spike protein and cultivated it in the Madin Darby Canine Kidney Cells [132]. The University of Mehmet Ali Aydunar in Turkey and Indian Immunologicals Ltd in India also have projects on attenuated COVID-19 vaccines, while none of these vaccine candidates have stepped into phase three clinical trials.

Inactivated whole-virus vaccine

Vaccines that use inactivated microorganisms to induce immune response also have a very long history. Although it is a whole (killed) virus, it does not have the risk of viral reactivation compared with live attenuated viruses. Moreover, this vaccine is effective on multiple SARS-CoV-2 antigens, including the S protein. Therefore, the inactivated virus is stable and safe compared to live attenuated viruses [136]. A phase 2 clinical trial only reported mild side effects, such as localized injection site redness and pain, while all had resolved within 72 h [137]. There are strong research foundations of inactivated SARS-CoV-2 because this strategy has already been tested for SARS-CoV and many other diseases, such as Cholera, Hepatitis A virus, Plague, and the like. However, several limitations made inactivated whole-virus vaccines less attractive. First, it needs a long time to produce a dose of this kind of vaccine, so that it is not appealing enough in the time-sensitive race of COVID-19 vaccine development [138]. Besides, whole-inactivated vaccines have to be manufactured in biosafety level three-capable facilities, which will also cause difficulties for some areas [139]. Second, the duration of immunity is short that demands inoculation of higher amounts of vaccine with an adjuvant [136]. Third, the inactivation process may also structurally deforms the immunogenic epitopes of inactivated viruses which can disable the vaccine [131].

Although the inactivated whole-virus has some disadvantages, the proper combination of the inactivation method and adjuvants can also produce a satisfactory preventive effect. Sinopharm and Sinovac choose alum as the adjuvant [139, 140]. Among those inactivated candidates, the most promising inactivated COVID-19 vaccine candidates are in China [138]: CoronaVac/PiCoVacc (Registry index: NCT04456595NCT04582344), Inactivated COVID-19 vaccine (Vero cells) (Registry index: ChiCTR2000034780), and BBIBP-CoV (Registry index: NCT04560881 ChiCTR2000034780) have all stepped into phase 3 clinical trials.

Viral vectored vaccine

Viral vectored vaccines are another type of live attenuated vaccine. Viral vectors that cannot replicate in human cells are inserted specific DNA to induce a strong antigen-specific immune response. Although some replication active vectors can also be exploited as viral vector vaccines, they will not trigger any other disease. Ura et al. suggested that this kind of vaccine is highly specific in delivering the genes to the target cells, highly efficient in gene transduction, and induces immunity [132, 141]. Because researchers are only exposed to viruses and DNA that do not cause diseases, the risk of laboratory infection is greatly reduced. Some drawbacks also exist in the process of viral vectored vaccines production. Because we need to insert only DNA into viral vectors, including cellular system optimization and contaminants exclusion, complicated technologies are needed to ensure the efficiency of viral vectors [142]. Moreover, pre-existing immunity on the viral vector may affect vaccine efficacy. For instance, the most widely used vectors for COVID-19 vaccines are adenoviral-based. Most patients may have pre-existing immunity against the adenovirus, and the immune response to SARS-CoV-2 may be dampened [143]. Last, there is a risk of genomic integration of recombinant viruses into the host which may lead to cancer, and additional long-term biosafety assessment is required.

Kaur et al. [132] suggested that viral vector vaccines had great potential for prophylactic use. There are some candidates in phase 3 clinical trials. ChAdOx1 (NCT04516746NCT04540393) was developed by the University of Oxford, and the Adenovirus vector genome is formatted by inserting the SARS-CoV-2 spike gene into the E1 locus of the ChAdOx1 adenovirus genome. Ad5-nCoV (NCT04526990NCT04540419) developed by the Beijing Ins of Biotechnology was constructed by the Admax system from the Microbix Biosystem [144]. Both RBD and spike protein-specific neutralizing antibody increases were observed within two weeks of immunization, whereas the pre-existing anti-Ad5 immunity would limit the immune response [144]. Beyond these two projects, Gam-COVID-Vac (NCT04530396NCT04564716) and Ad26.COV2⋅S (NCT04505722) also have the potential of being marketed within a short time.

Protein subunit vaccine

This kind of vaccine consists of viral proteins or protein fragments. Because it does not have any live component of the viral particle, it is one of the safest candidates with fewer and milder side effects. For the biosafety for a new virus causing a global pandemic, numerous vaccine projects are focusing on the SARS-CoV-2 proteins and their fragments. As a cost of safety, the subunit vaccine can only induce limited immunity and requires auxiliary adjuvant to activate a robust immune response. Most of the SARS-CoV-2 subunit vaccines have focused on the spike protein or its fragments, such as the RBD [145]. Because of the activation of antibodies and contains HLA-restricted T-cell epitopes in the studies of MERS and SRAS-CoV, the N protein has also drawn some attention [146, 147].

NVX-CoV2373, produced by Novavax, is the leading candidate for the SARS-CoV-2 protein vaccine. This vaccine uses a saponin-based adjuvant to increment the adaptive immune responses to recombinant Ebola virus glycoprotein vaccines and MVA-based influenza vaccines [148]. This candidate uses a nanoparticle with the full-length spike protein that was capable of binding ACE2 receptors with high affinity. Instead of using the full-length spike protein, another candidate produced by Anhui Zhifei Longcom (NCT4464085) only utilizes the RBD of the spike protein. The University of Queensland in collaboration with GSK and Dynavax intends to use the Molecular Clamp technology to produce the neutralizing antibodies and use the adjuvant platform technology to enhance the vaccine response [132].

DNA vaccine

DNA-based vaccination approach is said to be the most revolutionary approach that encodes for the antigen [132]. DNA-based vaccination approach can increase the stability of the DNA molecule so that it can be used to construct a large number of mRNA molecules. Moreover, because the DNA vaccine is stable at room temperature, there is no need for a cold chain for this kind of vaccine. There are few DNA vaccines registered for human use, and the pipelines of DNA-based candidates for SRAS-CoV-2 can be divided into two ways: naked DNA plasmids and naked DNA plasmids plus electroporation [136].

INO-4800, produced by Inovio Pharmaceuticals, is a prophylactic DNA vaccine against COVID-19. The vaccine was created based on the synthesized SRAS-CoV-2 IgE-spike sequence, and it was suggested that the vaccine can provide an effective immunity within a week after vaccination [149]. In the USA, there are more than three registered DNA vaccine candidates (https://clinicaltrials.gov/), but no additional information is available.

RNA vaccine

Although messenger RNA (mRNA) has not yet produced any registered vaccine in the past, the production of nucleic acid-based vaccine is often faster and cheaper than protein subunit vaccines [150], so that there have been six RNA vaccines for COVID-19 in the phase of clinical trials [134]. This kind of vaccine consists of viral antigen-encoding mRNAs that can be translated by human cells to produce antigenic proteins and induce immunity. Similar to DNA vaccines, RNA vaccines are easy to modify the antigenic proteins, while they will not interact with host-cell DNA and thus no genomic integration risk exists. The biggest challenge of RNA vaccines is the propensity of mRNA to degrade. The vaccine is not as stable as the DNA vaccine and it must be stored and handled carefully. In addition, there is a potential risk of RNA-induced interferon response which is associated with inflammation and autoimmunity [151].

The mRNA-1273 vaccine produced by Moderna in collaboration with the National Ins of Allergy and Infectious Diseases is the leading RND vaccine candidate of SRAS-CoV-2. It encodes the spike-2 protein antigen and they substituted the amino acids at 986 and 987 with proline to stabilize the protein [131]. Several studies suggested that the mRNA-1273 vaccine is safe and effective in preventing COVID-19 [152, 153].

Inequities among countries

Globally, to avoid the inequity in COVID-19 vaccine distribution among countries and to prevent a repeat of vaccine protectionism like the H1N1 scenario [86], the COVID-19 Vaccine Global Access (COVAX) Facility was established by WHO in April 2020 and was co-led by the Vaccine Alliance (Gavi), the Coalition for Epidemic Preparedness Innovations (CEPI). With lower prices for low-income and middle-income countries, COVAX aims to provide all countries the opportunity to get at least 20% population vaccinated at the end of 2021. Under its vision, no country can be the first to vaccinate over 20% of its population if other countries cannot reach this rate [164].

Unfortunately, COVAX has not achieved its goal yet. Although the global administered coverage of the second dose has reached 15.5% in 9th Aug 2021, the United Kingdom, France, and the USA have vaccinated for more than 40% of its population, while the rates in low- and middle-income countries like Zambia, Sudan, and Nigeria are even lower than 10% (https://ourworldindata.org/grapher/covid-vaccination-doses-per-capita). Because COVAX cannot make itself the only way to purchase vaccines, many high-income countries with smaller populations directly purchase vaccines from the developer instead of COVAX, which makes them host most of the vaccines in the world [165]. Moreover, insufficient funding has also prevented COVAX from providing an adequate share of vaccines to low- and middle-income countries that are more in need of them [166].

Besides the lack of a global vaccine stockpile, some researchers have also criticized the allocation strategies of COVAX [167, 168]. At the first stage in COVAX’s allocation structure, a proportional allocation strategy will be implemented for every country until all countries have reached the coverage goal. However, when high-income countries have broken the role on the distribution of vaccines, equity in population is no longer equal to fairness in need. Some additional allocation models based on ethical equity have been proposed as complements of COVAX, such as the fair priority model (FPM) [168] and multi-value ethical framework (MEF) [86].

Inequities among populations

When limited vaccines are available in the early stage of the vaccination program, policymakers have to decide on who should be vaccinated first. Accounting for both health and economic benefits, all countries face a multicriteria decision problem. As recommended by WHO, many countries have given priority to key workers, clinically vulnerable groups, and elderly groups under the structure of the Strategic Advisory Group of Experts on Immunization (SAGE) [169, 170]. However, pregnant women, an important vulnerable group, were excluded from the recruitment lists of most of the global clinical trials [171]. Without evidence about safety and efficacy during pregnancy, pregnant women have fallen into a perpetuated cycle of exclusion [172]. When pregnant women are not considered in early efforts of vaccine development, they are in turn excluded from generations of evidence, and they are excluded from vaccination programs. Ignoring pregnant women, who have equivalent or even worse outcomes compared with non-pregnant groups, are harmful to the global COVID-19 disease burden and are unjust in the progress of global immunization of COVID-19 vaccines.

Many researchers have advocated the inclusion of pregnant groups [172–174], and the Pregnancy Research Ethics for Vaccines, Epidemics, and New Technologies (PREVENT) Working Group even issued 22 recommendations to promote equity for pregnant women in vaccine development and response [175]. Although pregnant women were not in most of the cohorts of trials, some pregnancies occurred across the trials of three vaccines previously approved in the UK [176]. The data showed no significant detrimental effect of vaccination on pregnancy. Thus, the United Kingdom, the United States, and European Union all recommended pregnant people to be vaccinated if the benefits outweigh the potential risks. Moreover, Shimabukuro et al. [177] provide more scientific evidence about vaccine safety in pregnant persons. Based on a vaccination health checker surveillance system data of 35,691 participants in the USA, they find no obvious safety signals among pregnant persons who received mRNA COVID-19 vaccines. Although it is said that more longitudinal follow-ups with large numbers of women vaccinated earlier in pregnancy are needed, at least the mRNA vaccines of COVID-19 are most likely adequate for the pregnant group. As mentioned in the previous section, mRNA vaccines are not very stable and have high requirements for the cold chain. Due to the high cost of cold chain transportation, there is no well-operated cold chain network in many low-income countries. Even if mRNA vaccines are completely safe and effective for pregnant women, the inequity of pregnant women vaccination still exists in low-income countries.

Racial inequities in vaccination also persist. Take the United States as an example, from the latest data on COVID-19 vaccinations by race, nearly two-thirds of people who had received at least one dose of the vaccine were White (58%), and only 10% were Black and 6% for Asian [178]. Black people have only received a small share of vaccination, whereas more than half of cases and deaths were made up of them. Despite the hesitancy for the vaccine, a long time of medical racism and misinformation has caused the unfairness of vaccination for Black [179]. CDC in the USA has made many efforts to ensure that Black and other minorities have fair and just access to COVID-19 vaccination. But as the mantra of the COVID-19 pandemic, “No one is safe until everyone is safe” [180], the efforts of filling the gap in the vaccination of minorities are far from enough.