The BNT162b2 vaccine (Pfizer/BioNTech) has shown high efficacy1 in preventing COVID-19 as well as real-world effectiveness for prevention of infection, symptoms, hospitalization, severe disease and death2,3. Moreover, we and others have reported that the vaccine reduces viral load of BTIs, suggesting that it can also reduce infectiousness4,5. Indeed, national vaccination campaigns were followed by reduced infection rates, with the effect extending beyond the individual vaccinees to the community level6,7,8, and giving rise to hopes for herd immunity and disease eradication. In Israel, a rapid national vaccination campaign (over 50% inoculated within 3 months) was followed by near-complete eradication of new COVID cases in the period from April to June 2021 (ref. 9).

However, despite the comprehensive vaccination campaign and its initial dramatic effect, since mid-June 2021, a new surge has occurred in several highly vaccinated countries, including Israel. This increased infection rate, and concomitant increased rate of BTI, comes amid a global spread of the Delta variant of SARS-CoV-2 (B.1.617.2)10 and a possible waning of the vaccine-induced immune response, raising questions of vaccine effectiveness against the Delta variant over time. Considering the time since vaccination, studies have shown waning efficacy in protection against infection after 6 months11,12. Nevertheless, to date, only a slight decline has been observed in preventing severe disease and its consequences13. Considering the viral loads of BTIs, reports vary substantially as to whether, and to what extent, the vaccine is effective in reducing viral loads of BTI with the Delta variant14,15.

Given this uncertainty, to restrain the current wave of the pandemic, Israel, followed by the United States and the United Kingdom, has decided to offer and recommend a third dose (or booster) to people for whom 5 months have passed since their second dose. In Israel, the booster dose was first offered to people aged 60 years or older and then gradually extended to younger age groups. First reports of real-world data in humans indicate that the booster dose is effective in preventing infections16. Similarly, a booster study in primates against the Beta variant using the mRNA-1273 vaccine has shown up to a 20-fold increase in neutralizing antibodies17. However, it is still unclear how the booster shot affects the viral load in BTI in humans and in Delta-dominant real-world settings.

In this study, we retrospectively collected and analyzed the reverse transcription quantitative polymerase chain reaction (RT–qPCR) test measurements of three SARS-CoV-2 genes—E, N and RdRp (Allplex 2019-nCoV assay, Seegene)—from positive tests of patients at Maccabi Healthcare Services. We focus on infections of adults over age 20 between 28 June and 9 September 2021, when Delta was the dominant variant in Israel (over 93%)9. Crossing this dataset with vaccination data (59% of Israel’s population was two-dose vaccinated by 28 June), we identified, in total, 3,100 infections of unvaccinated, 12,934 BTIs of two-dose-vaccinated and 519 BTIs of booster-vaccinated individuals (Methods: ‘Vaccination status’; Table 1).

Table 1 Study population

Considering all of these infections (n = 16,553), we built a multivariable linear regression model for the cycle threshold (Ct) value of each of three SARS-CoV-2 genes, accounting for vaccination at different time bins before the infection and for receiving the booster, as well as adjusting for sex, age and calendric date. Focusing on the RdRp gene, regression coefficients for vaccinated over unvaccinated, representing the difference in Ct between the groups, started with 4.6 (95% confidence interval (CI), 2.2–6.9) for BTI 7–30 d after the second vaccine dose, yet decayed over time down to 0.6 (95% CI, 0.05–1.12) after about 2 months (P = 0.00005, Methods: ‘Change in Ct over time’), and vanished to insignificant values for infections 6 months or longer after vaccination. Analyzing data for up to 34 d after inoculation revealed that this decline, however, was overturned after the booster shot, which was associated with an increase of 2.4 (95% CI, 2.0–2.9) in the Ct. Noting that a difference of 1 Ct unit is approximately equivalent to a factor of 2 in the number of viral particles per sample, this increase in the Ct corresponds to a more than five-fold reduction in viral load (Methods: ‘Linear regression’; Fig. 1a). Similar changes in Ct were also observed for the two other genes, N and E (Extended Data Fig. 1). Although the model adjusts for sex and age, noting that, due to national rollout guidelines, the average age of the booster group was higher than that of the two-dose-vaccinated group (58.6 versus 42.0 years old), we also repeated the multivariable linear regression analysis separately for patients older and younger than age 50, which yielded similar results (Extended Data Fig. 2). Furthermore, as full protection might not always be fully realized 7 d after either the second dose or booster dose16, we repeated the analysis considering time bins starting 14 d after these doses and observed similar associations (Extended Data Fig. 3). To avoid possible effects of severe cases on the results, we also analyzed the data while excluding hospitalized patients, obtaining similar results (Extended Data Fig. 4). Finally, we note that, because during the current surge only a small fraction of patients is within their initial 2-month post-vaccination period, considering the entire population as a whole, there is only a very small viral load difference between the vaccinated and unvaccinated groups (0.22 (95% CI, 0.02–0.42); Fig. 1b).

Fig. 1: Association of infection Ct with two-dose vaccination and with the booster.
figure 1

a, Ct regression coefficients, indicating an infection Ct relative to unvaccinated control group (dashed line), show an initial increase in Ct in the first 2 months after the second vaccination dose, which then gradually diminishes, ultimately vanishing for infections occurring 6 months or longer after vaccination. Increased Ct is restored after the booster (right bar). Coefficients were obtained by multivariate linear regression analysis, adjusting for age and sex (Methods; n = 16,553). b, Same model as in a but without binning post-vaccination times. Because most of the vaccinated population during the current surge are more than 2 months after their second vaccine shot (Table 1), at the whole population level the average effect of the vaccine on Ct is negligible. Error bars represent one standard error. *P < 0.05, **P < 0.01, ***P < 0.001; all P values are two sided. Data are shown for Ct of the RdRp gene; for genes N and E, see Extended Data Fig. 1a, b. NS, not significant.

Source data

Our results show that the vaccine is initially effective in reducing viral loads of Delta BTIs, with a magnitude of ten-fold (95% CI, 4–30) (average over the first 2 months after vaccination), consistent with its initial effectiveness against pre-Delta variants4,5. However, this viral load reduction effectiveness declines with time after vaccination, significantly decreasing at 3 months after vaccination and effectively vanishing after about 6 months. As the Delta variant surfaced when a large fraction of the vaccinated population was already past the initial 2-month post-vaccination period, the population-wide average effect of the vaccine on Delta viral loads is negligible, consistent with and explaining reports14,15 of no difference in Ct between vaccinated and unvaccinated individuals infected with Delta. Consistently, we found that a booster shot is associated with a regained viral load reduction even during a Delta-dominated surge, suggesting restored vaccine-associated mitigation of transmissibility. In total, these results suggest that, although the initial vaccine effectiveness in reducing viral loads of Delta-variant BTIs is similar to its observed effectiveness in pre-Delta surges, this effectiveness is reduced with post-vaccination time and can then be restored with a third booster dose, consistent with immunity against infection waning and third inoculation boosting.

Our study had several limitations. First, although viral load is a common proxy for infectiousness, positive PCR results do not necessarily imply a viable virus, and a correlation between viral loads and infectiousness is not fully established. Second, recent studies suggested that a faster decline in viral load occurs in vaccinated compared to unvaccinated individuals with Delta infection14,15, offering a potential differentiation effect on Ct for samples taken long after the onset of symptoms. It is, therefore, also possible that the difference in Ct between vaccinated and unvaccinated individuals reflects not only an initial difference in viral loads between these two groups but also a differential viral load decay between the groups. However, we note that we minimized this effect by considering only the first positive test for each patient, which is typically performed within 5–6 d after infection, namely 1–2 d after symptoms16,18,19. Third, it is unknown how many of the infections in our dataset were symptomatic or asymptomatic, and the fraction of symptomatic infections in the unvaccinated population could be higher. Lastly, potential socio-behavioral differences between vaccinated and unvaccinated individuals might also affect the composition of our dataset.

It is yet to be seen for how long the booster’s renewed effect of reducing BTI viral loads will last and whether there will be a need for additional booster doses in the future against the same variant or others to come. Furthermore, vaccine effectiveness against severe infections, although correlated with viral loads20, might wane at different time scales and be affected differently by booster inoculation, as, to date, only a slight decline has been observed in preventing severe disease13. Nonetheless, at least in the short term, the lower viral load among booster-vaccinated Delta BTI points to lower infectiousness which, together with other means, such as social distancing and masks, could help impede the spread of the pandemic.


Data collection

Anonymized SARS-CoV-2 RT–qPCR Ct values were retrieved for all of the positive samples taken between 28 June and 9 September 2021 and tested at the Maccabi Healthcare Services central laboratory using SQL Server Management Studio, version 18.5.1. No statistical method was used to predetermine sample size. Patient data included sex, year of birth, any record of COVID-associated hospitalization and indication of immunosuppression. For patients with multiple positive tests, only the first positive test was included. Vaccination dates for these patients were retrieved from the centralized database of Maccabi Healthcare Services. Immunocompromised patients were excluded (n = 720). Patients were excluded if their first positive sample was between the first shot and 1 week after the second shot. Patients with a first positive sample within the first 1 week after the booster shot were also excluded. Patients with positive test results before the study period were also excluded. For each test, Ct values for E gene, RdRp gene, N gene and the internal control were determined using Seegene proprietary software for the Allplex 2019-nCoV assay after the standard oro-nasopharyngeal swab specimen collection procedure. The same PCR model was used for all of the tests (Bio-Rad CFX96 Real-Time PCR Detection System).

Vaccination status

Patients who tested positive 7 d or more after the second shot were regarded as ‘vaccinated’. Patients who tested positive 7 d or more after the third shot were regarded as ‘booster-vaccinated’. Patients who tested positive fewer than 7 d after the second shot were excluded, as were patients who tested positive fewer than 7 d after their booster shot.

Change in Ct over time

To calculate the change in Ct between post-vaccination time bins, we applied the same linear regression model as above, except replacing the length 5 one-hot vector of post-vaccination time with a length 3 time-accumulated binary vector indicating [0,0,0], [1,0,0], [1,1,0] and [1,1,1] for infections in unvaccinated or 7–60 d, 61–180 d and over 180 d after the second vaccination. The coefficient of the second binary variable, indicating the difference in Ct between infections in the first 2 months and infections 2–6 months after vaccination is −3.1 (95% CI, (−4.6) to (−1.6)), P = 0.00005 for the RdRp gene; −2.9 (95% CI, (−4.3) to (−1.4)), P = 0.0002 for the E gene; and −3.0 (95% CI, (−4.5) to (−1.4)), P = 0.0002 for the N gene.


Linear regression

For each of the three viral genes, we calculated the linear regression of Ct values as a function of time since the second shot (length 5 one-hot vector indicating 0/1 for time bins of 7–30 d (or 14–30 for Extended Data Fig. 3), 31–60 d, 61–120 d, 121–180 d and more than 180 d after the second shot; all 0s for unvaccinated), booster status (0/1), sex (0/1, female/male), age (bins of 10 years) and calendric date (number of days since 28 June 2021). Models were implemented using Python’s statsmodels library, version 0.9.0, and scipy library, version 1.1.0.

Ethics committee approval

This study was approved by the Maccabi Healthcare Services institutional review board (IRB). Owing to the retrospective design of the study, informed consent was waived by the IRB, and all identifying details of the participants were removed before computational analysis.

Reporting Summary

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