2.1 Study population, data sources, and study design

This study is divided into two parts:

1. A retrospective analysis dealing with the effectiveness of different homologous and heterologous vaccination schemes at various time points. Vaccines included in the analysis were as follows:
   1. Sinopharm (BBIBP-CorV)
   2. Pfizer (BNT162b2)
   3. Sputnik V (Gam-COVID-Vac)
   4. AstraZeneca (ChAdOx1 nCoV-19)
2. A prospective analysis dealing with the immunogenicity of different homologous and heterologous vaccination schemes at different time points.

2.2 VE against COVID-19 infection and hospitalization

This retrospective cohort case-control study examined the effectiveness of two-dose primary vaccination schemes (2× Sinopharm, 2× Pfizer, 2× Sputnik V, 2× AstraZeneca, or 1× Sinopharm/1× Pfizer) and homologous (2× Sinopharm/1× Sinopharm or 2× Pfizer/1× Pfizer) or heterologous (2× Sinopharm/1× Pfizer, 2× Sputnik V/1× Pfizer, or 2× Pfizer/1× Sinopharm) first booster vaccination schemes against laboratory-confirmed SARS-CoV-2 infections and subsequent hospitalization during the Alpha (B.1.1.7), Delta (B.1.617.2), and Omicron (B.1.1.529, with its sublineages BA.1, BA.2, BA.4, and BA.5) variants that circulated in Lebanon between 2021 and 2022, compared with unvaccinated individuals without previous exposure to COVID-19.

The study population comprised active Lebanese military personnel between February 2021 and September 2022, who were followed longitudinally throughout the study period. Data were retrieved using the Lebanese Ministry of Defense (MoD) COVID-19 electronic healthcare database. The study included personnel who had received the mentioned vaccination schemes as well as unvaccinated personnel, both with and without previous exposure to COVID-19. Personnel who were partially vaccinated or who received vaccination schemes other than those previously mentioned were excluded from the analysis. Data, including SARS-CoV-2 polymerase chain reaction (PCR) testing results; COVID-19 vaccination history; clinical infection and hospitalization; and other demographic and comorbidity information, including cardiovascular disease, pulmonary disease, kidney disease, diabetes mellitus, malignancy, and obesity, were extracted by the principal investigators from the digital health information database.

Since the onset of the pandemic, the MoD followed the Lebanese MoH recommendations and guidelines regarding SARS-CoV-2 PCR testing, contact tracing, isolation and quarantine, absenteeism, and management and treatment of infection, which were in line with the recommendations of the United States (US) Centers for Disease Control and Prevention (CDC) and the World Health Organization (WHO) [14–17]. In alignment with the national vaccination campaign, the MoD initiated COVID-19 vaccinations for military personnel in mid-February 2021 and began administering booster doses in November 2021. The MoD followed the Lebanese MoH’s guidance on COVID-19 vaccination schedules, which were based on the recommendations of the CDC and WHO [15, 17]. Personnel received vaccines at military treatment facilities and non-military vaccination centers. Regarding the type of vaccine administered, there was no preference, as all were donated to the Lebanese military institution.

A PCR-positive swab, irrespective of the reason for PCR testing or the presence of symptoms, was used to define laboratory-confirmed SARS-CoV-2 infections. Breakthrough SARS-CoV-2 infections were considered to occur at least 14 days after the second dose of the primary vaccination regimen listed above and at least 7 days after receiving the booster dose [18–20]. A positive SARS-CoV-2 PCR test within 3 weeks of a previous positive SARS-CoV-2 PCR test was excluded and not counted as a new infection. Infection severity classification followed the WHO guidelines for COVID-19 case severity (acute-care hospitalizations) [14].

In Lebanon, the COVID-19 pandemic was characterized by the dominance of three SARS-CoV-2 variants during 2021 and 2022, as determined by analyzing the Lebanese dataset in GISAID [21–24]:

1. The Alpha (B.1.1.7) variant predominantly circulated until the end of May 2021.
2. By the end of July 2021, the Delta (B.1.617.2) variant had completely replaced the Alpha (B.1.1.7) variant.
3. The Delta (B.1.617.2) variant, in turn, was gradually replaced by the Omicron (B.1.1.529, sublineages BA.1 and BA.1.1) variant in the second week of November 2021 until the end of December 2021.
4. The Omicron (B.1.1.529, sublineages BA.1, BA.1.1, or BA.2) variant predominantly circulated between January 2022 and May 2022.
5. By mid-June 2022, the Omicron variant of concern (B.1.1.529, sublineages BA.4 or BA.5) had completely replaced the previously mentioned sublineages.

Coincidentally, the appearance of each variant and/or sub-lineage was followed by new surges in the number of cases in Lebanon [21–24]. We chose to study VE during five periods of high or peak circulation of SARS-CoV-2 variants, excluding data from the early and late stages of the wave. In the initial stages, transmission is slower, and toward the end of the wave, herd immunity begins to develop (Fig 1). Personnel who tested positive were only included up to September 8, 2022, as it marked the end of the study period. Both symptomatic and asymptomatic laboratory-confirmed infections were included. This approach ensured that the same individuals were followed over time, allowing for a comprehensive assessment of VE across multiple periods.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. COVID-19 in Lebanon: The daily number of COVID-19 cases throughout the waves of the pandemic in Lebanon and associated SARS-CoV-2 variants between January 2021 and September 2022.

Reference: Edouard Mathieu, Hannah Ritchie, Lucas Rodés-Guirao, Cameron Appel, Charlie Giattino, Joe Hasell, Bobbie Macdonald, Saloni Dattani, Diana Beltekian, Esteban Ortiz-Ospina and Max Roser (2020)—"Coronavirus Pandemic (COVID-19)". Published online at OurWorldInData.org. Retrieved from: ’https://ourworldindata.org/coronavirus’ [Online Resource]. N.B.: X-axis: month-year; Y-axis left: number of confirmed daily cases. In this study, we examined vaccine effectiveness during five periods of high or peak circulation of SARS-CoV-2 variants [21–24]: 1. SARS-CoV-2 Alpha cases: Personnel were categorized as such if they initially tested positive between February 21st, 2021, and May 31st, 2021, as the majority of cases in Lebanon were primarily attributable to the circulating Alpha variant during that period. 2. SARS-CoV-2 Delta cases: Personnel were classified as such if they first tested positive between July 21st, 2021, and November 7th, 2021, as the majority of cases in Lebanon were predominantly attributable to the circulating Delta variant during that time. 3. SARS-CoV-2 Mixed (Delta and Omicron (B.1.1.529)) cases: Personnel fell into this category if they first tested positive between November 8th, 2021, and December 31st, 2021. This period represented an overlap between two variants, with residual Delta infection incidence in the community alongside the gradual increase in Omicron infections. 4. SARS-CoV-2 Omicron (B.1.1.529, sublineages BA.1, BA.1.1, or BA.2) cases: Personnel were designated as such if they first tested positive between January 1st, 2022, and March 7th, 2022, as these sub-lineages predominantly circulated during that time frame. 5. SARS-CoV-2 Omicron (B.1.1.529, sublineages BA.4 or BA.5) cases: Personnel were identified as such if they first tested positive between June 21st, 2022, and September 7th, 2022, as these sub-lineages predominantly circulated during that period.

https://doi.org/10.1371/journal.pone.0306457.g001

In each of the five periods, VE analyses were stratified according to primary immunization schemes (2× Sinopharm, 2× Pfizer, 2× Sputnik V, 2× AstraZeneca, or 1× Sinopharm/1× Pfizer). Cases and controls were categorized based on the elapsed time since the last vaccine dose. VE was assessed for each primary course in intervals of < 3 months, ≥ 3 –<6 months, and ≥6 months after the second dose. VE was assessed for the first booster (Sinopharm or Pfizer vaccine) in intervals of < 3 months, ≥ 3 –< 6 months, and ≥ 6 months after receiving the booster dose. VE was calculated for each vaccination scheme, including all vaccinees with and without previous exposure to COVID-19.

2.3 Humoral immunity

Between December 2022 and January 2023, we conducted a random prospective assessment of immunogenicity at various time points in vaccination subgroups that had previously received Sinopharm and Pfizer vaccines, and had not been exposed to COVID-19. Participants were categorized into different groups based on their primary vaccination type and whether they had received a booster dose (Table 1). The number of participants in each group during the immunogenicity assessment period was determined by the temporal definitions of the groups. Participants were further stratified by age and gender.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Immunogenicity assessment groups based on the primary and booster vaccination schemes and timeframe from the second or booster dose.

https://doi.org/10.1371/journal.pone.0306457.t001

2.4 Ethics approval

This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines (S1 Table). It was conducted in accordance with the guidelines of the Declaration of Helsinki, and ethics approval was obtained from Makassed General Hospital’s Institutional Clinical Research Ethics Committee (approval numbers: 1112021 and 1102022). The section of the study related to VE was deemed not to involve human participants due to the retrospective nature of the study; thus, the requirement for informed consent was waived. During the data collection phase, only subject case numbers were included. At a later stage, a different number was assigned to each case to safeguard subject privacy. All methods were performed in accordance with the hospital’s institutional review board committee guidelines and regulations. However, participants in the immunogenicity subgroup analyses provided their written informed consent for the collection of their health information and blood samples, with a focus on ensuring anonymity.

2.5 Statistical analysis

Sociodemographic cohort characteristics, including age categories (<50 and ≥50–<65 years), gender (male and female), COVID-19 infection/recovery history before each variant surge (yes and no), personnel function (healthcare worker and non-healthcare worker), and the presence of comorbidities (none, one comorbidity or more), were compared between vaccination groups and presented as numbers and percentages. Categorical analyses were performed using the chi-square test.

For VE against infection during each of the five periods of high or peak circulation of SARS-CoV-2 variants, multivariable logistic regression was used, with laboratory-confirmed SARS-CoV-2 infections as the dependent variable and case participants being those who tested positive. Controls included personnel with no record of a positive SARS-CoV-2 PCR test during each period, either due to a lack of clinical suspicion or any possibility of contracting COVID-19 based on the contact tracing system implemented by the MoD since the start of the pandemic, irrespective of their vaccination status.

For VE against hospitalization during each of the five previously mentioned periods, multivariable logistic regression was used, with acute-care hospitalization due to laboratory-confirmed SARS-CoV-2 infections as the dependent variable, and case participants being those hospitalized. Controls included personnel with no record of acute-care hospitalization due to laboratory-confirmed SARS-CoV-2 infections during each period, as documented in the MoD healthcare database, irrespective of their vaccination status.

For VE against infection and hospitalization calculation, vaccination status was included as the independent variable, and the effectiveness of different vaccination schemes over the five time periods and the associated 95% confidence interval (CI) were calculated using the following equation: VE = [1 − adjusted odds ratio (aOR) of vaccination among cases compared with controls] × 100. VE was adjusted in logistic regression models for age categories, gender, COVID-19 infection/recovery history before each variant surge, personnel function, and the presence of comorbidities. These factors were all considered potential confounders and were included in all models. VE was calculated in the entire cohort, with the reference group for all estimates being the unvaccinated COVID-19-naïve personnel (i.e., individuals with no previous infection) in each of the five periods. Only those vaccinated in this specific time-since-vaccination stratum and those unvaccinated in each VE (infection) analysis for a specific time-since-vaccination stratum were included. Consequently, the number of cases and controls varied across the time-since-vaccination analyses.

Regarding immunogenicity subgroup analysis, categorical analyses on age and gender, presented as numbers and percentages, were performed using the chi-square test for different vaccination schemes at different time points among the selected participants. Antibody titers were presented as geometric mean titers (GMT) and 95% CIs. One-sample Kolmogorov–Smirnov test was used to check for data distribution normality. Kruskal–Wallis test, followed by Dunn’s multiple comparison post-hoc test, was performed to compare unpaired nonparametric data between the groups (antibody levels).

Statistical significance was defined as p < 0.05. The IBM Statistical Package for the Social Sciences program for Windows (version 23.0) (Armonk, NY, USA: IBM Corp.) and GraphPad Prism 9.0 software (GraphPad Software, Inc., San Diego, CA, USA) were used to conduct statistical analyses using two-tailed tests.

3.1 Characteristics of the study population for VE analyses

The total number of active military personnel identified in the COVID-19 healthcare database of the Lebanese MoD and longitudinally evaluated for VE across different time periods was 83,760. The distribution of vaccinees based on different vaccination schemes and non-vaccinated personnel during the study period is illustrated in Fig 2. Baseline demographics and characteristics of the study cohort according to the different vaccination combination groups are provided in S2 Table.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Study profile.

Abbreviations: VE = vaccine effectiveness, SPh = Sinopharm (BBIBP-CorV), AZ = AstraZeneca (ChAdOx1 nCoV-19), PFZ = Pfizer-BioNTech (BNT162b2), SPTV = Sputnik V (Gam-COVID-Vac).

https://doi.org/10.1371/journal.pone.0306457.g002

3.2 VE against COVID-19 infection across various periods and time points

3.2.1 Alpha variant infection.

3.2.1.1 At less than 3 months after the second dose. The effectiveness of Sinopharm primary vaccination scheme against infection was highest at 99.8% (95% CI, 99.6–99.9), similar to Pfizer primary vaccination scheme which reached 98.8% (95% CI, 97.84–99.36) (see Fig 3A and Table 2). No assessment of VE change with time against the Alpha variant infection was possible for any primary scheme during this wave.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

Effectiveness of the available COVID-19 vaccination schemes against laboratory-confirmed infections during the (A) Alpha variant wave, (B) Delta variant wave, (C) Mixed Delta and Omicron BA.1 variant wave,(D) BA.1/BA.2 Omicron variant wave, and (E) BA.4/BA.5 Omicron variant wave. Abbreviations: AZ = AstraZeneca (ChAdOx1 nCoV-19), mo = months, PFZ = Pfizer-BioNTech (BNT162b2), SPh = Sinopharm (BBIBP-CorV), SPTV = Gamaleya’s Sputnik V (Gam-COVID-Vac), VE = vaccine effectiveness. N.B. Data are presented as effectiveness point estimates, with error bars indicating the corresponding 95% confidence intervals.

https://doi.org/10.1371/journal.pone.0306457.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Effectiveness of the available COVID-19 primary and booster vaccination schemes against laboratory-confirmed infections during the different waves.

https://doi.org/10.1371/journal.pone.0306457.t002

3.3 VE against acute-care hospitalization due to COVID-19 infection across various periods and time points

3.3.1 Alpha variant infection-related acute-care hospitalization.

3.3.1.1 At less than 3 months after the second dose. Both the Sinopharm and Pfizer primary vaccination schemes demonstrated high effectiveness against hospitalization at less than 3 months following the second dose, with estimates of 98.6% (95% CI, 95.6–99.8) and 96.4% (95% CI, 83.0–99.8), respectively.

No significant change in effectiveness was observed for any primary scheme during this period.

Additionally, no critical or fatal COVID-19 cases were reported among breakthrough infections. (see Fig 4A and Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4.

Effectiveness of the available COVID-19 vaccination schemes against acute-care hospitalization attributed to laboratory-confirmed infections during the (A) Alpha variant wave, (B) Delta variant wave, (C) Mixed Delta and Omicron BA.1 variant wave, and (D) BA.1/BA.2 Omicron variant wave. Abbreviations: AZ = AstraZeneca (ChAdOx1 nCoV-19), mo = months, PFZ = Pfizer-BioNTech (BNT162b2), SPh = Sinopharm (BBIBP-CorV), SPTV = Gamaleya’s Sputnik V (Gam-COVID-Vac), VE = vaccine effectiveness. N.B. Data are presented as effectiveness point estimates, with error bars indicating the corresponding 95% confidence intervals.

https://doi.org/10.1371/journal.pone.0306457.g004

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Effectiveness of the available COVID-19 primary and booster vaccination schemes against acute-care hospitalization attributed to laboratory-confirmed infections during the different waves.

https://doi.org/10.1371/journal.pone.0306457.t003

3.4 Assessment of humoral immunity of different vaccination schemes in the subgroup analysis

The majority of participants included in this subgroup analysis, who received various vaccination schemes, were predominantly males under 50 years of age and had no comorbidities. Demographic characteristics of the participants recruited for the immunogenicity analysis are presented in S7 Table. Results of the anti-S-IgG GMT (BAU/ml) in the groups are presented in Fig 5A–5C and Table 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Humoral immune responses to SARS-CoV-2 spike protein following various primary vaccination schemes and booster doses with Sinopharm or Pfizer vaccines among COVID-19-naïve participants at different time points.

Abbreviations: IgG = immunoglobulin G, BAU = Binding Antibody Unit, mo = months, PFZ = Pfizer-BioNTech (BNT162b2), SPh = Sinopharm (BBIBP-CorV). N.B: The initial measurement for anti-S-IgG was deemed negative if the index was < 1.00 (seronegative) and positive if the index was ≥ 1.00 (seropositive). All readings were standardized to BAU/mL using the WHO international standard for the VIDAS®3 SARS-CoV-2 IgG (VIDAS®3 SARS-CoV-2 IgG index = 1 (cutoff) = 20.33 BAU/mL). Antibody titers were reported as geometric mean titers (GMT) with corresponding 95% confidence intervals (CIs). One-sample Kolmogorov–Smirnov test was used to assess data distribution normality. Kruskal–Wallis test, followed by Dunn’s multiple comparison post-hoc test, was used to compare unpaired nonparametric data among the groups (antibody levels). ns (not significant), P > 0.05, *: P ≤ 0.05, ***: P ≤ 0.001, ****: P ≤ 0.0001.

https://doi.org/10.1371/journal.pone.0306457.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Immunogenicity results of different vaccination schemes at different time points.

https://doi.org/10.1371/journal.pone.0306457.t004

• The initial measurement for anti-S-IgG was deemed negative if the index was < 1.00 (seronegative) and positive if the index was ≥ 1.00 (seropositive).
• All readings were standardized to BAU/mL using the WHO international standard for the VIDAS®3 SARS-CoV-2 IgG (VIDAS®3 SARS-CoV-2 IgG index = 1 (cutoff) = 20.33 BAU/mL).

4.1 Study scope and cohort characteristics

After 4 years of the COVID-19 pandemic, we have gained a comprehensive understanding of vaccine dynamics [17, 28–30]. Our study provided a unique opportunity to compare the effectiveness of different vaccines administered to the same population simultaneously, a scenario that is uncommon in many settings. Typically, countries rely on a single vaccine type or platform during specific periods as part of their national vaccination programs. For instance, the US and most European countries predominantly used mRNA vaccines, while China relied on inactivated vaccines [17, 28–30].

In contrast to controlled trials and laboratory studies that assess immunogenicity or efficacy, which are conducted in controlled environments with specific timeframes, subjects, and locations, VE measurement relies on real-world data [13, 31]. Real-world effectiveness of vaccines is influenced by numerous factors, including individual immune responses, adherence to vaccination schedules, and community-level influences [13, 31]. In addition, VE can vary even for the same vaccine within the same population, depending on when the effectiveness was assessed during a variant wave [19]. As previously mentioned, at the time of the outbreak, transmission is slower, and toward the end of the wave, herd immunity builds up, potentially leading to a false negative impact on calculated VE. This effect could mask the true effectiveness of the vaccine when calculated. Given these considerations, we chose to focus solely on the peak of the variant wave, excluding data from the early and late stages of the wave.

Our study cohort exhibited relatively homogenous demographic and other characteristics. It primarily comprised young, healthy, nonhealthcare worker males, with a median age of 32 years (interquartile range, 27–38 years). Vaccine responsiveness and severity of breakthrough infections are influenced by factors such as age, immune conditions, and comorbidities [32]. To minimize the impact of these confounding variables in more diverse populations, we selected a relatively homogeneous cohort of generally healthy individuals for immunogenicity and effectiveness studies. However, the effects of the vaccine on patients with different clinical characteristics should be explored in separate, well-defined cohorts. This approach offers a more comprehensive perspective by examining various facets of VE and mitigating the influence of confounding factors.

4.2 VE against documented infection across the different vaccination schemes and variants

In this case-control study, the effectiveness of primary COVID-19 vaccine schemes, including Sinopharm, Pfizer, and Sputnik V, remained consistently high during the Alpha and Delta waves, despite a decline in the latter. However, during the Omicron wave, significant declines in VE against infection were observed across all vaccine platforms. This decline began during the phase when “Mixed” Delta and Omicron variants were present between November and December 2021 and persisted until the Omicron variants predominated in 2022. Primary vaccination schemes involving Sinopharm, Pfizer, Sputnik V, and AstraZeneca lost their protective effect against infection. Nonetheless, a heterologous primary vaccination scheme (1× Sinopharm/1× Pfizer) demonstrated a noteworthy protective effect. It is important to interpret this result with caution, as the number of individuals who received this scheme was relatively small (n = 197) compared with the other groups, although the difference reached statistical significance.

These findings align with several real-world effectiveness studies conducted globally that have consistently documented substantially lower protection by the different vaccines against infection with the Delta and Omicron variants compared with previous variants such as Alpha or the wild-type SARS-CoV-2 strain [28, 29, 33, 34]. COVID-19 vaccine efficacy was outpaced by viral mutations, explaining why various vaccines were highly effective against earlier variants but less so against Delta and Omicron variants. The emergence of these variants led to multiple waves, both locally and globally [35].

For instance, a nationwide Hungarian study evaluating VE during the Alpha wave found adjusted effectiveness against infection as follows: Pfizer: 83.3%; Sputnik V: 85.7%; AstraZeneca: 71.5%; and Sinopharm: 68.7% [33]. Similarly, a study from the UK observed high VE during the Alpha wave for Pfizer’s and AstraZeneca’s primary vaccination schemes but noted declines during the Delta wave [34].

In a nationwide Danish cohort study, mRNA vaccines were found to be highly effective against Delta variant infection but less so against Omicron, with VE estimates of 92% and 40%, respectively, during the first month after vaccination, in individuals aged < 60 years [29]. Similar results were reported in Belgium, where Pfizer’s VE was 96% against Alpha, 87% against Delta, and 31% against Omicron [28].

A recent systematic review and meta-analysis of studies on inactivated SARS-CoV-2 vaccines’ real-world effectiveness against infection revealed a pooled VE estimate of 53% against the Delta variant and a lower estimate of 16% against the Omicron variant, which circulated during 2021 and 2022 [36], consistent with our findings.

Early concerns about SARS-CoV-2 evolution focused on the potential emergence of variants with significant antigenic differences that could evade immunity acquired through vaccination or previous infection [37, 38]. Most widely utilized COVID-19 vaccines, which employ various platforms, as listed above, were initially developed based on the spike protein from early virus variants or the original wild-type virus [39].

While the Alpha variant exhibited limited antigenic changes, the Delta variant had a moderate ability to evade vaccine-induced antibodies [37, 38, 40, 41]. However, VE estimates remained acceptable against Delta [37, 38, 40, 41]. The Omicron variant, with its numerous antigenic sub-lineages, possesses a higher degree of mutation and significant antigenic drift, rendering it less neutralized by first-generation vaccines [37, 38, 40, 41]. Real-world data on VE against infections reflect these antigenic changes [37, 38, 40, 41].

In our cohort, booster doses were administered mostly 6 months after the second dose, coinciding with the transition from “Mixed” variants to the predominance of Omicron. VE generally increased, providing higher protection within three months after receiving a Pfizer booster. This trend was observed across primary vaccination schemes, but boosting with Sinopharm did not protect against infection in those initially vaccinated with Pfizer’s and Sinopharm’s primary schemes during Omicron phases. These results suggest that boosting is more effective with mRNA vaccines.

Consistent with our findings, a national study conducted in Malaysia compared the effectiveness of homologous and heterologous boosters. The results revealed that homologous inactivated vaccine boosting (VE = 33%) and adenovirus vaccine boosting (VE = 30%) were less effective than heterologous boosting (VE = 48%) and homologous boosting (VE = 51%) using Pfizer during the predominant Omicron period [42]. In addition, a previous study in Lebanon during the first Omicron wave showed similar effectiveness of mRNA boosters against infection, with VE measured at 64% and 57% for individuals who had previously received Pfizer’s and Sputnik V’s primary vaccination schemes, respectively [19].

4.3 Effect of time on VE against documented infection across the different vaccination schemes and variants

As previously discussed, vaccine-mediated protection against COVID-19 infection diminishes over time, particularly with the emergence of diverse SARS-CoV-2 variants such as Delta and Omicron. Consequently, booster doses have been introduced to strengthen and prolong immunity, with updated versions of COVID-19 vaccines targeting Omicron subvariants. In our cohort, changes in VE were evaluated from the Delta wave onwards. During the Delta wave and the phase characterized by “Mixed” variants, we observed fluctuations in VE estimates for Sinopharm’s and Pfizer’s primary vaccination schemes. VE declined between 3 to less than 6 months after the second dose but remained above 50%. However, VE of all primary schemes declined more rapidly during the Omicron phases, dropping below 20% at 3 to less than 6 months after the second dose and reaching null values at 6 months and beyond. Assessment of booster doses’ protection against infection was only possible during the second Omicron wave (BA.4/BA.5), which began in June 2022. The decline of booster VE with time was similarly steep for all homologous and heterologous booster schemes, with a slightly steeper decline observed for the homologous boosters from 3 to more than 6 months after the booster and beyond.

These findings align with real-world effectiveness reports consistently showing a significant decline in immunity against infection over time, with the most notable decrease observed with the Omicron variant compared to earlier variants such as Alpha or Delta [28, 29, 36]. In the previously mentioned pooled analysis of VE estimates of the primary schemes of inactivated vaccines, protection against SARS-CoV-2 infection decreased significantly after 6 months following primary vaccination, reaching 21% during the Delta wave and 7% during the Omicron wave [36]. For primary schemes of mRNA vaccines, the aforementioned Danish study reported waning immunity of the primary schemes across several variants, with effectiveness at 73.2% for Alpha, 50% for Delta, and 4% for Omicron, at more than 4 months since vaccination [29].

Regarding booster shots with inactivated vaccines for individuals who had received inactivated vaccines as their primary series, Kyaw and colleagues reported a complete loss of protection against infection just 2 months after the booster dose in a pooled analysis of 6 real-world effectiveness studies against the Omicron variant [43]. However, they also noted a less steep decline in VE against Omicron for BNT162b2 heterologous boosters in individuals who had initially received inactivated vaccines as their primary series. VE decreased from 57% in less than 3 months after the booster receipt to 35% thereafter [43].

Notably, the effectiveness estimate for the Sputnik V primary scheme increased from negative values within less than 3 months from the second dose to 57% within the 3 to less than 6 months following the second dose, particularly during the Omicron BA.1/BA.2 wave. Similarly, during the Omicron BA.4/BA.5 wave, VE estimates for the 2× Sputnik V/1× Pfizer regimen intriguingly rebounded from negative values within 3–<6 months after the booster to a remarkable 91% at 6 months and beyond. Dolzhikova and colleagues analyzed neutralizing antibody responses against variants of concern in sera samples of individuals vaccinated with Sputnik V, including those who were revaccinated with Sputnik Light [44]. Their study revealed that antibodies triggered by the Sputnik V vaccination matured and developed a broader capacity to neutralize emerging variants of concern within 6–9 months after the initial vaccination and 2–3 months after receiving a booster [44]. Similarly, a longitudinal study from Argentina demonstrated robust SARS-CoV-2 neutralizing antibodies and reduced viral variant escape to neutralization over time, particularly 6 months after primary vaccination with Sputnik V [45]. Investigators reported that antibodies produced in individuals vaccinated with Sputnik V exhibited increased cross-neutralization capacity against variants of concern as time passed [45]. Martynova and colleagues demonstrated that Sputnik V vaccination elicits a robust antibody response, recognizing diverse epitopes on the S-protein and a strong cellular response. These responses remain detectable for more than 3 months postvaccination, indicating the vaccine’s long-term efficacy [46].

4.4 VE against acute-care hospitalization

The significance of VE against hospitalization becomes apparent when VE estimates against infection for specific variants with particular vaccine protocols decline. Overall, our results revealed that VE estimates against hospitalization for primary vaccination schemes were generally lower during the Delta wave compared with the Alpha wave. However, all primary schemes remained highly protective, with effectiveness above 80% against hospitalization across all postvaccination timeframes. During the phase characterized by the presence of “Mixed” Delta and Omicron variants and continuing thereafter during the Omicron-predominant waves, sharp declines in effectiveness against acute-care hospitalization were observed for primary vaccination schemes. Booster schemes provided full protection against hospitalization during the “Mixed” phase of the Delta and Omicron variants, but a decline was noted in booster protection when the Omicron BA.1/BA.2 variant predominated. Notably, all primary and booster vaccination schemes provided 100% protection against critical illness and mortality during all waves in our cohort. Contrary to the common belief that effectiveness against infection wanes over time while effectiveness against hospitalization lasts, our data has shown that effectiveness against hospitalization also declines with time across all primary and booster schemes, especially during Omicron predominance.

The time-sensitive waning of VE against hospitalization was observed against Delta infection-induced hospitalization, but it is dramatically accelerated against Omicron infection-induced hospitalization [47]. In an observational prospective case-control study involving 21 hospitals in the US, Pfizer’s primary series was effective against both Alpha and Delta variant-associated hospitalization, with VE measuring 85% [48]. However, lower protection was observed against Omicron-associated hospitalization, reaching 65% [48]. mRNA boosters increased effectiveness against Delta and Omicron-induced hospitalizations to 94% and 86%, respectively [48]. In a Brazilian nationwide case-control study, the effectiveness of the CoronaVac primary series against COVID-19 associated hospitalization was substantially lower during an Omicron-dominated period (56%) compared with a Delta-dominated period (87%) [49]. During the Omicron-dominated period, a homologous CoronaVac booster dose conferred a smaller increase in protection against severe disease (74%) compared with a heterologous Pfizer booster (86%) [49]. Notably, the increased protection afforded by a homologous booster against severe disease waned during the 4 months after its administration, in contrast to the durable effectiveness of the heterologous Pfizer booster [49].

VE against infection exhibited statistically significant negative values within both primary and booster vaccination schemes during the periods when the Omicron variant predominated. However, it is crucial to note that these negative estimated values of effectiveness likely reflect the influence of certain factors rather than indicating a true negative biological effectiveness of the vaccines [50, 51]. These factors include vaccinated individuals having a higher risk of network-level exposure, increased social contact rates, greater susceptibility, lower rates of prior infection, or less adherence to safety measures and barriers such as mask usage, hand hygiene, and social distancing compared with unvaccinated individuals [50, 51]. In addition, over time, the number of vaccinated individuals increased, along with those previously exposed to the virus, leaving unvaccinated, COVID-naïve adults as a minority, especially during later waves of the pandemic [52–54]. These factors significantly impact vaccine’s effectiveness against recent variants like Omicron compared with earlier variants.

Based on our data, which demonstrates that vaccine protection against critical illness and death in a generally young and healthy population remains preserved over time, frequent booster shots may not be necessary for this group, particularly those who can tolerate infection without severe complications. This is especially pertinent as VE against infection tends to diminish with time and with the emergence of new variants. We propose that young and healthy individuals, such as those included in this cohort, may not require regular vaccine boosters once they have established protection against critical illness and death through primary vaccination. Their immunity can potentially be naturally reinforced, akin to the immune response historically seen with common colds before the COVID-19 pandemic. However, this conclusion may not apply to more vulnerable individuals, such as those with comorbidities or immune deficiencies that impede the initial vaccine response. Separate studies should focus on these individuals to assess VE against infection, hospitalization, critical illness, or COVID-19-associated mortality. This will enable evidence-based decisions regarding the optimal vaccine regimens and booster schedules.

4.5 Immunogenicity of vaccination schemes involving Sinopharm and Pfizer

In primary vaccination schemes, the homologous Pfizer regimen elicited a higher humoral response compared to Sinopharm. However, the humoral response generated by the homologous Pfizer regimen was comparable to that of the heterologous primary regimen 1× Sinopharm/1× Pfizer. Over time, we observed a decline in antispike IgG GMT for both Sinopharm’s and Pfizer’s primary vaccination schemes. Pfizer’s booster dose elicited a similarly robust humoral response in individuals who initially received either Pfizer’s or Sinopharm’s primary vaccination schemes. However, a Sinopharm booster dose induced a significantly lower humoral response compared to the latter booster schemes in individuals who had initially received either Pfizer or Sinopharm as their primary vaccination.

These findings are consistent with results from other immunogenicity studies in the literature, where antispike IgG GMT levels were consistently higher among recipients of the primary Pfizer vaccine scheme compared to those who received Sinopharm [55, 56]. Humoral immunity was observed to be short-lived and tended to decline over time, as documented in other studies [57–59]. Heterologous boosting has been explored in individuals who previously received primary inactivated vaccine schemes. In a prospective cohort study evaluating the immunogenicity of Pfizer’s booster among Peruvian healthcare workers who had previously received the primary Pfizer or Sinopharm vaccination schemes, investigators reported that the antispike IgG GMT levels produced by the heterologous vaccine regimen was significantly higher than that elicited by the homologous booster regimen [56].

The concept of an immune correlate of protection (or immuno-bridging) aims to utilize immunological measurements as predictive markers for protection against infectious diseases [37]. Before the emergence of variants that could evade vaccine efficacy, serology testing and the stratification of seroconversion levels proved useful in promptly identifying high-risk groups of vaccine non-responders. These groups may not develop a viral neutralizing response, even if they show seroconversion, and therefore could remain at a higher risk of infection despite vaccination [60].

Throughout the pandemic, studies have consistently indicated a strong relationship between high titers of antispike RBD IgG and a less severe disease state, as well as a lower incidence of breakthrough infections. Higher IgG binding has been associated with increased protection against infection with the wild-type, Alpha, and Delta variant strains of the SARS-CoV-2 virus [61–64]. Furthermore, a robust correlation has been observed between antispike RBD IgG titers and neutralizing antibody titers, suggesting that IgG concentrations provide insights into protection against infection with the ancestral strain or relatively homologous variants of concern [64–66].

In our study, Pfizer vaccine elicited a significantly higher immune response compared to Sinopharm after initial vaccination. Nevertheless, both vaccines demonstrated high levels of protection against infection caused by the Alpha and Delta variants, but their effectiveness gradually decreased and diverged from each other when it came to the Omicron variants.

Consistent with our findings, a study from Thailand by Takheaw and colleagues determined the correlations between the levels of anti-RBD IgG and neutralizing antibodies against SARS-CoV-2 variants in vaccinated individuals with two doses of CoronaVac or one dose each of CoronaVac and AstraZeneca [67]. The investigators reported a high correlation between anti-RBD IgG and neutralizing antibodies against the wild-type, Alpha, and Delta variants but not with the Omicron variant [67]. Among individuals with high levels of anti-RBD IgG, 93% had neutralizing antibodies against the wild-type, Alpha, and Delta variants, but none had neutralizing antibodies against Omicron [67]. The study concluded that anti-RBD IgG levels cannot predict the presence of neutralizing antibodies against the Omicron variant [67].

In a study from Sweden, Vikström and colleagues found no indication that levels of vaccine-induced antibodies after an mRNA booster dose protected against infection with the Omicron variant [68]. They found no pattern supporting that higher levels of circulating S-binding IgG protected against SARS-CoV-2 infection [68]. They attributed their findings to the Omicron variant’s significant escape from neutralizing antibodies [68].

4.6 Limitations

This study, relying on real-time data, encountered several limitations. First, there were instances of missing data, particularly those regarding the progression of VE against specific variants over time within defined time points and for particular vaccination schemes. This limitation arose from the retrospective collection of data during different waves of the pandemic, with the follow-up period coinciding with subsequent surge in cases related to other emerging variants.

Second, a relatively young and healthy cohort was studied, which may restrict the generalizability of our findings to the entire Lebanese population. However, this demographic characteristic can also be viewed as a strength, as it allows for the study of effectiveness while controlling for confounding factors such as age and comorbidities.

The detection of laboratory-confirmed infections relied on SARS-CoV-2 PCR testing of symptomatic and asymptomatic individuals during contact tracing. Ideally, systematic screening at multiple time intervals would have allowed us to detect more asymptomatic infections. Nevertheless, our study was based on real-world data in a low-income country with limited resources, facing a socioeconomic crisis. The bias toward including only symptomatic infections rather than all infections was minimized by the vigilant contact tracing system established by the MoD since the start of the pandemic. In addition, the definition of variant-specific cases was based on epidemiologic evidence rather than genomic testing of the variants in the studied population.

The negative estimated effectiveness values during the Omicron waves were more likely a result of bias rather than a genuine indication of the vaccines having negative biological effectiveness. This bias may have arose from factors such as the gradual reduction of the unvaccinated population without prior COVID-19 exposure over time and the increased exposure of vaccinated individuals to the virus. In addition, there may be unequal compliance with safety measures when compared with the unvaccinated population. These factors can lead to an underestimation of VE [69].

It is important to note that the sample size for VE against COVID-19 acute-care hospitalization during Omicron periods was limited in this study. While every effort was made to include all eligible cases, the small sample size during these periods may affect the generalizability of our findings. Thus, caution should be exercised when interpreting the results from these time periods.

The lack of cellular immunity and serum-neutralizing antibody testing in the immunogenicity evaluation was a limitation. Notably, humoral immunity was not assessed for vaccination schemes including Sputnik V or AstraZeneca vaccines. Furthermore, assessing changes in humoral response beyond 3 months post-first booster in any scheme was not possible.

Despite these limitations, our study offers a comprehensive view of real-world effectiveness for different homologous and heterologous primary and booster vaccination schemes against various vaccine-tolerant variants, specifically during periods of high viral circulation rather than throughout the entire wave.