## Main

Appreciable community transmission of the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) Delta (B.1.617.2) variant was first noted in Qatar by end of March 2021 (refs. 1,2,3). Although Delta incidence has increased along with a recent surge in cases and hovered at about 200 cases per day in the summer of 2021, it remains low compared to earlier variant incidences with no signal for an epidemic wave materializing as of 19 September 2021. Between 23 March 2021 and 7 September 2021, 43% of diagnosed infections were Delta infections (Methods)1,3. Delta dominance was, however, preceded by two large consecutive SARS-CoV-2 Alpha (B.1.1.7) and Beta (B.1.351) waves earlier in 2021 (refs. 1,2,3,4,5). The rapid scale-up of Coronavirus Disease 2019 (COVID-19) vaccination in Qatar may have impeded efficient Delta transmission. As of 19 September 2021, it is estimated that over 80% of Qatar’s resident population has received two doses of either the BNT162b2 (ref. 6) (Pfizer-BioNTech) vaccine or the mRNA-1273 (ref. 7) (Moderna) vaccine8. This study assessed BNT162b2 and mRNA-1273 vaccines’ real-world effectiveness against the Delta variant in Qatar from 23 March 2021 to 7 September 2021 and compared these estimates to those in other countries.

## Results

### Study population

From 21 December 2020 to 7 September 2021, 950,232 people had at least one BNT162b2 vaccine dose (median date of first dose was 21 April 2021) and 916,290 were fully vaccinated (median date of second dose was 11 May 2021). Administration of the second dose was within a median of 21 d after the first dose (interquartile range (IQR) 21–22 d), with full-vaccination of 97.4% of individuals within 30 d of first dose.

Over this timeframe, 564,468 individuals had at least one mRNA-1273 vaccine dose (median date of first dose was 19 May 2021) and 509,322 were fully vaccinated (median date of second dose was 24 May 2021); distributions for both doses were skewed with means of 16 May 2021 and 11 June 2021, respectively. Administration of the second dose was within a median of 28 d after the first dose (IQR 28–31 d), with full-vaccination of 74.7% of individuals within 30 d of the first dose.

With greater and regular vaccine availability, coverage for BNT162b2 has been steadily increasing since December 2020. In contrast, coverage for mRNA-1273 depended on dispatch of large shipments and did not reach considerable levels before March 2021.

We defined a Delta ‘case’ as a PCR-positive swab with the Delta variant, irrespective of the reason for the PCR test or symptom presence or absence (Methods). Infections with other variants were excluded, except for Beta in an additional analysis. All records of vaccination for both BNT162b2 and mRNA-1273 were included. Extended Data Figs. 13 show flowcharts depicting the selection of study populations to estimate effectiveness of BNT162b2 (Extended Data Fig. 1), mRNA-1273 vaccine (Extended Data Fig. 2) and either of these vaccines (Extended Data Fig. 3) against the Delta variant. Tables 1 and 2 describe the samples used in estimation of effectiveness ≥14 d after the first dose and ≥14 d after the second dose, respectively. The median age of participants ranged from 26–30 years; only 9% of Qatar’s residents are ≥50 years of age and 89% are residents from more than 150 countries9,10.

### Delta vaccine-breakthrough infections

Delta cases were ascertained using real-time PCR with reverse transcription (RT–qPCR) genotyping of randomly collected clinical samples (Methods)1,3. There were 88 and 1,126 Delta breakthrough infections between 23 March 2021 and 7 September 2021 among vaccinated individuals with one or two BNT162b2 doses, respectively and 60 and 187 Delta breakthrough infections among vaccinated individuals with one or two mRNA-1273 doses, respectively.

Additionally, by 7 September 2021, there were 4 and 15 severe Delta COVID-19 cases (acute care hospitalizations11; Methods) among vaccinated individuals with one or two BNT162b2 doses, respectively and 3 and 1 severe disease cases among vaccinated individuals with one or two mRNA-1273 doses, respectively.

Furthermore, there were one and two critical Delta COVID-19 cases (intensive care unit (ICU) hospitalization11; Methods) among vaccinated individuals with one or two BNT162b2 doses, respectively. The critical disease case reported after only one BNT162b2 vaccine dose also led to COVID-19 death (COVID-19 deaths12; Methods). There were no critical or fatal COVID-19 cases among those vaccinated with mRNA-1273.

### Effectiveness ≥14 d after the first vaccine dose

Effectiveness against Delta infection ≥14 d after only one dose was estimated at 45.3% (95% confidence interval (CI), 22.0–61.6%) for BNT162b2, 73.7% (95% CI, 58.1–83.5%) for mRNA-1273 and 58.0% (95% CI, 44.4–68.2%) for either of these vaccines (Table 3).

Effectiveness against any Delta-induced severe11, critical11 or fatal12 COVID-19 disease (Methods), 14 or more days after only one dose, ranged between 80–87% for BNT162b2, mRNA-1273 and either of these vaccines, but 95% confidence intervals were wide given the relatively small number of Delta disease cases (Table 3).

### Effectiveness ≥14 d after the second vaccine dose

Effectiveness against Delta infection 14 or more days after the second dose was 51.9% (95% CI, 47.0–56.4%) for BNT162b2, 73.1% (95% CI, 67.5–77.8%) for mRNA-1273 and 55.5% (95% CI, 51.2–59.4%) for either of these vaccines (Table 3).

Effectiveness against any Delta-induced severe11, critical11 or fatal12 COVID-19 disease 14 or more days after the second dose was 93.4% (95% CI, 85.4–97.0%) for BNT162b2, 96.1% (95% CI, 71.6–99.5%) for mRNA-1273 and 93.6% (95% CI, 85.9–97.1%) for either of these vaccines (Table 3).

Sensitivity analyses adjusting for previous infection and health worker status in conditional logistic regression analysis confirmed the main findings (Table 4).

Vaccine effectiveness against Delta infection for those ≥50 years of age was lower than that for those <50 for both vaccines (Supplementary Table 1). However, this result should be seen in the context that those ≥50 years of age received their second dose earlier than those <50. The median date of second vaccine dose for those ≥50 years of age was 9 April 2021, but was 19 May 2021 for those <50 years.

Effectiveness against symptomatic Delta infection 14 or more days after the second dose was estimated at 44.4% (95% CI, 37.0–50.9%) for BNT162b2, 73.9% (95% CI, 65.9–79.9%) for mRNA-1273 and 49.2% (95% CI, 42.8–54.9%) for either of these vaccines (Table 5). Symptomatic infection was defined as a PCR-positive swab collected based on clinical suspicion (symptoms indicative of a respiratory tract infection).

Effectiveness against asymptomatic Delta infection 14 or more days after the second dose was estimated at 46.0% (95% CI, 32.3–56.9%) for BNT162b2, 53.6% (95% CI, 33.4–67.6%) for mRNA-1273 and 45.9% (95% CI, 33.3–56.1%) for either of these vaccines (Table 5). Asymptomatic infection was defined as a PCR-positive swab collected in the absence of reported respiratory tract symptoms, such as during a survey or a random testing campaign (data sources in Methods).

For comparison, vaccine effectiveness against Beta infection was also estimated over the same period 23 March 2021 to 7 September 2021. Beta cases were also ascertained using RT–qPCR genotyping of randomly collected clinical samples (Methods)1,3. Effectiveness against Beta infection was estimated for BNT162b2 at 18.9% (95% CI, −1.8–35.4%) 14 or more days after only one dose and at 74.3% (95% CI, 70.3–77.7%) 14 or more days after the second dose (Table 6). The corresponding effectiveness measures for mRNA-1273 were 66.3% (95% CI, 55.8–74.2%) and 80.8% (95% CI, 69.0–88.2%), respectively. Estimated effectiveness against any Beta-induced severe11, critical11 or fatal12 COVID-19 disease was >90% for both vaccines (Table 6).

In comparing estimates for Beta to those for Delta, it must be noted that the median PCR diagnosis date was 15 April 2021 for Beta cases, but was 2 August 2021 for Delta cases. Beta dominated transmission earlier in the study, whereas Delta dominated transmission later in the study1,2,3,4,5. From 1 August 2021 to 7 September 2021, 83.6% of the RT–qPCR-genotyped cases were Delta cases (Methods).

## Discussion

BNT162b2 and mRNA-1273 vaccines both showed robust effectiveness (≥90%) against Delta-related hospitalization and fatality, in line with studies from the United Kingdom13,14, United States15,16,17,18 and Israel19. Despite many breakthrough infections, particularly for BNT162b2, there were limited instances of severe or critical disease among vaccinated individuals. In BNT162b2 fully vaccinated individuals, only 15 severe disease cases, 2 critical disease cases and 1 COVID-19 death were due to Delta. For mRNA-1273, only 1 severe disease case and no critical or fatal disease cases were reported.

Notably, estimated BNT162b2 or mRNA-1273 effectiveness against Delta infection 14 or more days after the first dose or 14 or more days after the second dose, were comparable. Recent evidence pointed to considerable waning of vaccine effectiveness over time, particularly for BNT162b2 (refs. 14,20,21,22,23). The high effectiveness against Alpha and Beta in Qatar in our previous studies (≥75%)4,5,24,25 as well as against Beta in this study (Table 6) were estimated when most residents in Qatar were recently vaccinated with BNT162b2 or mRNA-1273. Conversely, effectiveness against Delta was estimated here after several months have passed since the second vaccine dose for a large proportion of residents. This unexpectedly low effectiveness against Delta in fully vaccinated individuals could be therefore reflecting gradual waning of vaccine protection.

This observation is consistent with the pattern seen in reported effectiveness estimates against Delta elsewhere. Our estimate of 51.9% in BNT162b2 fully vaccinated individuals is lower than that reported in the United Kingdom14,26,27 and Canada28, where effectiveness was estimated at >75%, but similar to that reported in Israel19 and the United States18,29,30,31, where effectiveness was estimated between 39% and 66%. The delay in administering the second dose in the United Kingdom and Canada led to most persons being fully vaccinated ~3 months more recently than in Israel, the United States and Qatar, where vaccinated persons received their second dose 3 weeks after the first dose. The lower effectiveness in Israel, the United States and Qatar may therefore signal waning of vaccine protection in those who were fully vaccinated by the end of 2020 or early in 2021, as also suggested in a recent analysis of waning of BNT162b2 protection over time in Qatar23. Notably, mass vaccination in Qatar started shortly after that in Israel and the United States.

Another potential explanation pertains to the gradual easing of public health restrictions in Qatar in the last few months, at a time when Delta incidence has been slowly increasing. With more restrictions eased based on vaccination status, which is implemented through a mandatory mobile app (the Ehteraz app), vaccinated individuals may have had higher social contact rates than unvaccinated persons and may have adhered less strictly to safety measures, such as masks, due to their perception of lower risk32,33,34. Such risk compensation may even increase over time after completing the second dose, resulting in further normalization of behavior33,34,35. Vaccinated persons may therefore have higher risk of exposure to the virus than unvaccinated individuals, leading to increased infection incidence among those vaccinated, thereby reducing the observed real-world vaccine effectiveness.

Higher effectiveness against infection with Delta after the second dose was estimated for mRNA-1273 compared to BNT162b2 (P = 0.009), in line with studies indicating a stronger induced immune response and protection for mRNA-1273 (refs. 5,36,37,38).

This study found higher vaccine effectiveness for more serious COVID-19 disease (greater protection against symptomatic or severe infections), as observed earlier for BNT162b2 and mRNA-1273 effectiveness against the Alpha and Beta variants4,5,24,28.

This study has limitations. With the relatively small number of severe and critical disease cases and fatal cases in Qatar’s young population9,39, some of the effectiveness estimates against hospitalization and death had wide 95% confidence intervals. Data on comorbid conditions were not available to be included in the analysis. With the young population of Qatar9,10, the part of the population with serious comorbid conditions is small. In the national list of vaccine prioritization, there were only 19,800 individuals of all age groups with serious comorbid conditions. Accordingly, our findings may not apply to settings where the elderly population constitutes a considerable part of the population.

Data on occupation were not available to study investigators. The matching by nationality may have controlled in part for the occupational risk, considering the labor force structure in Qatar40,41,42. Infection incidence and vaccination were broadly distributed across the country’s neighborhoods or areas and population social substrata. Therefore, it is not likely that the results could be explained by clustering of vaccination or infection in specific geographies or social strata.

Vaccine effectiveness was investigated using a test-negative case–control study design43,44, rather than a randomized clinical trial design or a cohort study design that followed vaccinated and unvaccinated cohorts. However, the cohort study design applied to the same population of Qatar previously resulted in similar findings to the test-negative case–control study design4,5,45 (Extended Data Fig. 4), supporting the reliability of the test-negative case–control study design that has been of wide application for vaccine effectiveness studies of respiratory tract infections43,44.

In conclusion, both the BNT162b2 and mRNA-1273 vaccines are highly effective in preventing hospitalization and death due to infection with the Delta variant. However, effectiveness against infection was considerably lower than that against serious COVID-19 disease, particularly for the BNT162b2 vaccine. The reasons for the inferior protection against infection remain to be determined and may not necessarily relate to immune evasion by the Delta variant. The lower effectiveness may reflect some waning of vaccine protection over time23 or higher risk of exposure to the virus among vaccinated individuals compared to unvaccinated individuals, due to higher social contact rate and less adherence to safety measures. These findings indicate the need for more follow-up of vaccinated cohorts to investigate waning of vaccine immunity and for studies that investigate the effect of risk compensation on biasing vaccine effectiveness estimates.

## Methods

Hamad Medical Corporation and Weill Cornell Medicine-Qatar Institutional Review Boards approved the study with waiver of informed consent. A STROBE checklist is included in Supplementary Table 2.

### Data sources, study population and study design

This study was conducted in the resident population of Qatar. COVID-19 laboratory testing, vaccination, clinical infection data and related demographic details were extracted from the integrated, nationwide, digital-health information platform at Hamad Medical Corporation, the main public healthcare provider and the nationally designated provider for all COVID-19 healthcare needs. This platform hosts the national, federated SARS-CoV-2 databases. Data access was provided by the Ministry of Public Health for analyses to inform the national COVID-19 response. These databases include complete information for PCR testing, vaccinations, hospitalizations and demographic characteristics from epidemic onset.

Almost all vaccinations were provided at no cost in Qatar rather than abroad, through the universal public healthcare system for all nationals and residents of Qatar. In occasional episodes of vaccination abroad, details were still incorporated into the health system upon arrival to Qatar (at airport), for compliance with national regulations and to take advantage of travel-related privileges, such as quarantine exemption25.

All PCR tests in Qatar, irrespective of test-center location, are classified with respect to symptoms and the reason for testing (clinical symptoms, contact tracing, surveys or random testing campaigns, individual requests, routine healthcare testing, pre-travel, at port of entry or other). Only 9% of residents of Qatar are aged ≥50 years and 89% are incomers from over 150 countries9,10. Most of these expatriates are male craft and manual workers9,40,41.

We estimated vaccine effectiveness using a test-negative, case–control study design, a widely used design for appraising influenza vaccine effectiveness43,44. This design controls for potential bias due to infection misclassification or to healthcare-seeking differentials between vaccinated and unvaccinated individuals43,44. To maximize statistical power, all cases (PCR-positive individuals with confirmed SARS-CoV-2 Delta infection) and controls (PCR-negative individuals) in Qatar, between 23 March 2021 and 7 September 2021, were included in the study.

To adjust for underlying differences in the risk of exposure to infection9,40,41,42, we exact-matched cases and controls in a one-to-five ratio by sex, 5-year age group, nationality, reason for PCR testing and calendar week of PCR test. By virtue of having many more PCR-negative tests than PCR-positive tests, it was generally possible to find exact PCR-negative matches for most age groups for the PCR-positive Delta cases included in this study.

For each case, we considered the first PCR-positive test with confirmed Delta infection during the study from 23 March 2021 to 7 September 2021. After excluding all other PCR tests on individuals with infection, we considered the first PCR-negative test for each control during this period. This yielded an independent sample of unique cases and controls. This strategy was used to control for potential bias due to repeat testing in PCR-positive individuals seeking to check for infection clearance or bias arising from repeat testers among controls (persons with a higher level of healthcare-seeking behavior and presumably lower risk of infection).

PCR tests conducted for pre-travel or at the port of entry were excluded from analysis. This type of testing could possibly be affected by different test-seeking behavior among those vaccinated versus unvaccinated individuals given travel-related benefits extended only to vaccinated individuals, such as exemption from quarantine25.

We estimated effectiveness against Delta (B.1.617.2) documented infection (defined as a PCR-positive test with the Delta variant irrespective of the reason for the test or presence of symptoms) and against related severe, critical or fatal disease. Classification of case severity (acute care hospitalizations)11, criticality (ICU hospitalizations)11 and fatality12 was per WHO classification using individual chart reviews (details below).

We reviewed all PCR testing records for vaccinated and unvaccinated individuals. We excluded individuals with mixed vaccinations or with a vaccine record other than BNT162b2 or mRNA-1273. Every Delta case fulfilling the inclusion criteria, regardless of vaccination status and that could be matched to one or more controls was retained for the analysis. Infection and vaccination statuses were both ascertained at the time of PCR test. Each hospitalized individual underwent an infection severity assessment every 3 d from hospital admission up to discharge or death. Hospitalized individuals were classified according to their worst outcome (death12), followed by critical disease11 and severe disease11 (details below).

### COVID-19 severity, criticality and fatality classification

WHO defines severe COVID-19 as a SARS-CoV-2-infected individual with ‘oxygen saturation of <90% on room air and/or respiratory rate of >30 breaths min−1 in adults and children >5 years old (or ≥ 60 breaths min−1 in children <2 months old or ≥50 breaths min−1 in children 2–11 months old or ≥40 breaths min−1 in children 1–5 years old) and/or signs of severe respiratory distress (accessory muscle use and inability to complete full sentences and, in children, very severe chest wall indrawing, grunting, central cyanosis or presence of any other general danger signs)’11. Detailed criteria are in the WHO technical report11.

Critical COVID-19 is defined as a SARS-CoV-2-infected individual with ‘acute respiratory distress syndrome, sepsis, septic shock or other conditions that would normally require the provision of life sustaining therapies such as mechanical ventilation (invasive or noninvasive) or vasopressor therapy’11. Detailed criteria are in the WHO technical report11.

COVID-19 death is defined as ‘a death resulting from a clinically compatible illness, in a probable or confirmed COVID-19 case, unless there is a clear alternative cause of death that cannot be related to COVID-19 disease (for example, trauma). There should be no period of complete recovery from COVID-19 between illness and death. A death due to COVID-19 may not be attributed to another disease (such as cancer) and should be counted independently of preexisting conditions that are suspected of triggering a severe course of COVID-19’. Detailed criteria are in the WHO technical report12.

### Laboratory methods

Nasopharyngeal and/or oropharyngeal swabs were collected for PCR testing and placed in Universal Transport Medium (UTM). Aliquots of UTM were extracted on a QIAsymphony platform (QIAGEN) and tested with real-time RT–qPCR using TaqPath COVID-19 Combo kits (Thermo Fisher Scientific) on an ABI 7500 FAST (Thermo Fisher); tested directly on the Cepheid GeneXpert system using the Xpert Xpress SARS-CoV-2 (Cepheid); or loaded directly into a Roche cobas 6800 system and assayed with a cobas SARS-CoV-2 Test (Roche). The first assay targets the viral S, N and ORF1ab gene regions. The second targets the viral N and E-gene regions and the third targets the ORF1ab and E-gene regions.

Tests were performed at the HMC Central Laboratory or Sidra Medicine Laboratory, following standardized protocols.

### Classification of infections by variant type

Viral genome sequencing and multiplex RT–qPCR were used to screen for variants46 in randomly collected positive clinical samples1,2,3,4,5, supplemented by deep wastewater sequencing1,47. The latter is used to compare the distribution of variants in wastewater to that in clinical samples collected from patients with SARS-CoV-2.

Ascertainment of Delta (B.1.617.2) and Beta (B.1.351) cases in this study was through weekly RT–qPCR genotyping of positive clinical samples1,3. From 23 March 2021 to 7 September 2021, RT–qPCR genotyping identified 6,005 (35.5%) Beta (B.1.351)-like cases, 3,658 (21.6%) Alpha (B.1.1.7)-like cases, 7,218 (42.6%) ‘other’ variant cases and 51 (0.3%) B.1.375-like or B.1.258-like cases in 16,932 randomly collected specimens1,3. Since RT–qPCR genotyping started on 23 March 2021, the proportion of all diagnosed infections in Qatar that have been RT–qPCR genotyped is 12.0%, with the proportion of infections genotyped increasing with time, especially in the summer of 2021.

RT–qPCR genotyping accuracy was contrasted against results of Sanger sequencing of the receptor-binding domain of SARS-CoV-2 surface glycoprotein (S) gene or by viral whole-genome sequencing on a Nanopore GridION sequencing device. From 236 random samples (27 Alpha-like, 186 Beta-like and 23 ‘other’ variants), PCR genotyping results for Alpha-like, Beta-like and ‘other’ variants were in 88.8% (23 out of 27), 99.5% (185 out of 186) and 100% (23 out of 23) agreement with the SARS-CoV-2 lineages assigned by sequencing.

Within the ‘other’ variant category, Sanger sequencing and/or Illumina sequencing of the receptor-binding domain of SARS-CoV-2 spike gene on 728 random samples, between 23 March 2021 and 7 September 2021, confirmed that 701 (96.3%) were Delta cases and 17 (2.3%) were other variant cases, with 10 (1.4%) samples failing lineage assignment.6,8 Consequently, a Delta infection was proxied as any ‘other’ case based on the RT–qPCR-based variant screening result.

### Statistical analysis

Study samples were described using frequency distributions and measures of central tendency. The odds ratio (and 95% CI, comparing odds of vaccination among cases to that among controls), was estimated using conditional logistic regression factoring the matching in the study design. This analytical approach was implemented to reduce potential bias due to variation in epidemic phase43,48, gradual vaccination roll-out43,48 and other confounders9,40,41,42,49,50. CIs did not factor multiplicity. Interactions were not examined. Vaccine effectiveness at different time frames and its associated 95% CI were then estimated using43,44:

$${\rm{Vaccine}}\,{\rm{effectiveness}}=1-{\rm{odds}}\,{\rm{ratio}}\,{\rm{of}}\,{\rm{vaccination}}\,{\rm{among}}\,{\rm{cases}}\,{\rm{versus}}\,{\rm{controls}}$$

In each time-since-vaccination stratum, for first and second doses, we analyzed only those vaccinated in this specific time-since-vaccination stratum and those unvaccinated (our reference group). Accordingly, the sample size for cases (and controls) varied in the different time-since-vaccination analyses. As we used a test-negative study design, some individuals were tested PCR-positive or PCR-negative after their first dose and before the second dose. This allowed us to estimate effectiveness after only the first vaccination dose.

A sensitivity analysis was implemented to control for previous infection and health worker status in the conditional logistic regression, because health workers are potentially at higher risk of infection exposure and were prioritized for vaccination.

Additional analyses were performed to estimate vaccine effectiveness stratified by age (<50 versus ≥50 years of age). We also estimated vaccine effectiveness against symptomatic infection, defined as a PCR-positive swab collected based on clinical suspicion (symptoms indicative of a respiratory tract infection) and against asymptomatic infection, defined as a PCR-positive swab collected in the absence of reported respiratory tract infection symptoms (during a survey or a random testing campaign). For comparison, vaccine effectiveness was further estimated against the Beta variant, the only other variant with an appreciable incidence concurrent with the Delta incidence1,2,3.

A two-sided P value derived from logistic regression analyses was used to compare effectiveness of both vaccines with P < 0.05 showing statistical significance. Statistical analyses were conducted in STATA/SE version 17.0 (ref. 51).

### Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.