SARS-CoV-2 Vaccine Breakthrough by Omicron and Delta Variants, New York, USA

Discussion

Our exploration of vaccine breakthrough, vaccination status, and time since vaccination in this matched case–control study adds to the body of evidence supporting immune escape of SARS-CoV-2. Some results may seem counterintuitive because of the study design. For example, although a booster increases protection against infection with Omicron compared with absence of a booster,^[13,26] history of a booster was associated with Omicron (the emergent strain) and not Delta (the established strain) infection. This finding is consistent with evidence that suggests that having a booster is less effective for preventing infection with Omicron than with Delta.^[6,13] Similarly, vaccine effectiveness has been shown to wane with time;^[11] therefore, we hypothesized that increased time after vaccination would decrease the odds of being infected with the emergent strain.

Our analysis of New York state genomic surveillance data yielded results that are consistent with previous research showing an increased probability of breakthrough for Omicron compared with other variants for both vaccinated and boosted persons.^[6,8] In a similar study in Connecticut, USA, comparing odds of infection with Omicron versus Delta,^[6] an OR of ≈2 (95% CI 1.5–3.7 or 1.5–2.2, depending on time after vaccination) was found for vaccinated persons and ≈3 (95% CI 1.8–4.9) for boosted persons. These estimates are lower than the estimates from our study of 3.1 (95% CI 2.0–4.9) for vaccinated persons and 6.7 (95% CI 3.4–13.0) for unvaccinated persons, but the 95% CIs overlap between the 2 studies. A strong pattern of the emergent strain shows increased ability for vaccine breakthrough compared with other strains circulating at the time. Studies of prior variants of concern have found significant vaccine breakthrough in emergent variants. For example, Kustin et al. found that vaccine breakthrough for Alpha (B.1.1.7) was more likely compared with prior strains.^[27] Similarly, Tartof et al. found evidence for increased rates of vaccine breakthrough by Delta (B.1.617.2), although waning vaccine immunity was also a factor in that study.^[11] In addition, Rosenberg et al. showed increased breakthrough during the Delta emergence period and suggested that this effect was independent of waning immunity.^[28]

When we restricted the analysis to vaccinated persons only, time after vaccination was a statistically significant factor; probability of Omicron infection decreased with increased time after vaccination. The time-after-vaccination variable combined persons who had recently received a booster with those who had recently completed their primary series. Adding a variable to indicate booster status did not improve the model fit (Appendix Table 1). Of note, most persons in this study were >3 months past completion of their initial vaccination series. Boosters were more recent, and therefore vaccination status and booster status probably encoded much of the same information as a time-after-last-dose variable. No time-after-vaccination effect was detected if the data were coarsely divided into persons who had and had not received boosters, suggesting that more examination of this variable may be necessary. This variable was not found among the top models in the Delta emergence analysis.

Younger persons were more likely to be infected with Omicron than with Delta during the Omicron emergence period, although the data in this study cannot be used to distinguish a physiological basis from a behavioral basis for these age effects. Kahn et al. found Delta and Omicron infection be equally distributed by age among unvaccinated persons but to shift strongly toward younger persons among vaccinated persons;^[14] however, Accorsi et al. found elevated rates of Omicron infection among vaccinated and unvaccinated persons.^[13] It is possible that the age group effects are the result of a greater degree of socialization and other behavioral risk factors among persons 18–29 years of age. In 2020, college campus re-openings were associated with increased transmission of SARS-CoV-2.^[29] Because Omicron infections can break through vaccinations, college campuses may have increased the likelihood of persons in this age group being infected with SARS-CoV-2.^[30] The age group effect for preschool children (0–4 years of age) may represent a reduced level of socialization for this group. This effect, although included in the top model identified by the information theoretic approach here, was not statistically significant, so it also may be an artifact of low sample sizes for this age group. Other research has found that vaccines were not equally effective among age groups (V. Dorabawila, unpub. data, https://www.medrxiv.org/content/10.1101/2022.02.25.22271454v1). Vaccine effectiveness in New York was very low for persons 5–11 years of age, who received a lower dose (10 μg) of the Pfizer vaccine than for vaccinated persons ≥12 years of age who received a 30-μg dose (V. Dorabawila, unpub. data, https://www.medrxiv.org/content/10.1101/2022.02.25.22271454v1). However, the log-linear age effect detected here was not driven by children <12 years of age. When children <12 years of age were removed from the analysis, the estimated OR changed from 0.962 to 0.957 (95% CI 0.944–0.971), suggesting that the magnitude of the effect is greater when young children were removed from the analysis. Larger estimates for vaccination status and booster status were also greater when children <12 years of age were removed from the analysis (vaccination status OR 5.4, 95% CI 3.1–9.7; vaccination plus booster status OR 43.0, 95% CI 17.1–108.5). Vaccination rates and booster rates changed substantially during the study periods as well (Appendix Figures 2, 3), but any resulting biases were probably controlled for by the case–control study design.

Sample sizes were generally too small to detect robust vaccine type effects. The Janssen vaccine showed borderline significantly reduced OR for infection with Omicron relative to the Pfizer vaccine in 1 statistical model (Table 2; Appendix Table 1). This result would be consistent with improved performance against Omicron infection or with worse performance of this vaccine against Delta infection, as has been observed.^[28] Otherwise, OR estimates showed the potential for substantial differences, but overlapping 95% CIs prevent drawing robust conclusions (Table 2; Appendix Table 1).

Statistical power was constrained by the limited emergence periods and the relatively small percentage of viruses from COVID-19 case-patients that were sequenced. For Delta, the emergence period occurred during a time of reduced sequencing, because of low overall incidence during the summer of 2021, when Delta displaced previous strains (Figure 1). For Omicron, a larger sequencing effort was made, but the emergence period was considerably shorter because of the rapid dominance of the Omicron variant (Figure 1). Sample sizes could potentially be increased by expanding the regional scope of the study or incorporating sequencing results from other research laboratories.

We used only 1 matched set for each analysis. However, because case-patients were randomly matched to controls, other matches were possible. This limitation could be overcome by assessing significance with Monte Carlo simulation over the range of possible matches. That said, visual examination of leverage plots based on removing a single pair suggested that the results were generally unlikely to change with the removal of any single data point. The exception is the Delta analysis, in which a change of 1–2 data points would change the overall statistical significance of the results (Appendix Figure 1) without much change in the estimated OR.

In conclusion, this analysis of the emergence of the Omicron and Delta variants in New York, USA, based on sequenced virus identity broadly supports the results of prior studies.^[5–8] Vaccines offered less protection against Omicron infection, thereby increasing the number of potential hosts for emerging variants.

Emerging Infectious Diseases. 2022;28(10):1990-1998. © 2022 Centers for Disease Control and Prevention (CDC)