Multiple BCG vaccinations for the prevention of COVID-19 and other infectious diseases in type 1 diabetes

Abstract

There is a need for safe and effective platform vaccines to protect against coronavirus disease 2019 (COVID-19) and other infectious diseases. In this randomized, double-blinded, placebo-controlled phase 2/3 trial, we evaluate the safety and efficacy of a multi-dose Bacillus Calmette-Guérin (BCG) vaccine for the prevention of COVID-19 and other infectious disease in a COVID-19-unvaccinated, at-risk-community-based cohort. The at-risk population is made of up of adults with type 1 diabetes. We enrolled 144 subjects and randomized 96 to BCG and 48 to placebo. There were no dropouts over the 15-month trial. A cumulative incidence of 12.5% of placebo-treated and 1% of BCG-treated participants meets criteria for confirmed COVID-19, yielding an efficacy of 92%. The BCG group also displayed fewer infectious disease symptoms and lesser severity and fewer infectious disease events per patient, including COVID-19. There were no BCG-related systemic adverse events. BCG's broad-based infection protection suggests that it may provide platform protection against new severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants and other pathogens.

Trial registration: ClinicalTrials.gov NCT02081326.

Keywords: BCG; Bacillus Calmette-Guérin; COVID-19; Clinicaltrials.gov: NCT02081326; autoimmune; host microbe interactions; hygiene hypothesis; infectious diseases; phase 2/3 trial; randomized double blinded clinical trial; type 1 diabetes; vaccine.

Graphical abstract

Figure 1

Flow diagram and characteristics of participants (A) Flow diagram representing all enrolled participants from January 1, 2020 to April, 2021 for this double-blinded, randomized clinical trial testing repeat Tokyo-172 BCG vaccination versus placebo for COVID-19 protection. All 144 subjects were followed for 15 months with a 2:1 randomization and no dropouts. Data collection for this trial ended on April 2, 2021, the date when subjects started to receive provisionally approved COVID-19-specific vaccines. The geographical locations of the participants within the United States are shown in Figure S5. (B) Table of participant characteristics.

Figure 2

BCG vaccine efficacy and diagnostic confirmation of COVID-19 (A) Shown is the cumulative incidence of confirmed COVID-19 as a primary endpoint. During the 15-month surveillance time, one BCG recipient out of 96 (1.0%) met our criteria for molecularly confirmed COVID-19. In contrast, 6 out of 48 placebo recipients met the criteria (12.5%). Fisher’s exact testing showed a significant difference (two-tailed p = 0.006). Calculated vaccine efficacy was 92%, and the posterior probability (vaccine efficacy >30%) was 0.99. This was calculated using the Monte Carlo method. Vaccine efficacy was defined as (p1 – p2)/p1 × 100, where p1 is the percentage of COVID-positive subjects in the placebo group and p2 is the percentage of COVID-positive subjects in the BCG group. Our criteria for confirmed COVID-19 (see main text) required a combination of COVID-19 symptom(s) and ≥5 of 10 positive antibody assays including PCR, if available. Since many current clinical trials only define confirmed COVID-19 by symptom(s) and positive PCR testing, the PCR-only group was studied separately for vaccine efficacy (A and B). (B) Cumulative findings from each of the 10 diagnostic tests used to confirm COVID-19 (along with positive symptoms). These tests included the presence of COVID-19-specific antibodies to various SARS-CoV-2 virus epitopes through protein display (I–VIII), antibodies to the receptor-binding domain with an ELISA test (IX), and point-of-care testing (X) that included PCR. Our criteria for having confirmed COVID-19 required at least 5 of 10 detection methods to be positive, along with symptom(s). For the antibody assays, a patient was considered positive when the test resulted in a Z score of ≥3. In the cumulative graphs, the x axis data show the time period of the 15-month trial. The y axis shows the cumulative percentage of positive subjects. Except for the point-of-care graph (X), all other graphs represent the percentage of BCG and placebo patients with a Z score of ≥3 for the anti-SARS-CoV-2 antibody binding to a given protein region of the virus, i.e., the average antibody level during the COVID trial period was at least 3 standard deviations greater than the average level in the period preceding the COVID trial (baseline). The percentiles at the top right of each graph represent the calculated vaccine efficacy if this test alone was used to diagnose COVID-19 disease. Respective efficacy and Fisher’s exact p values for each COVID-19 antibody test were as follows: (I) 100%, p = 0.0007; (II) 80%, p = 0.089; (III) 91.7%, p = 0.009; (IV) 93.8%, p = 0.001; (V) 90%, p = 0.024; (VI) 87.5%, p = 0.004; (VII) 94.4%, p = 0.0003; (VIII) 83.3%, p = 0.009, (IX) 90%, p = 0.009; and (X) 91.7%, p = 0.029. For all graphs, BCG n = 96, placebo n = 48. Figure S1 shows the viral protein regions for each anti-COVID antibody tested. Number at risk data for each cumulative graph are shown in Figure S4.

Figure 3

Reduced COVID-19 disease markers after BCG vaccination (A) The heatmaps show Z scores for antibodies against SARS-CoV-2 in BCG- and placebo-treated symptomatic subjects (BCG n = 26, placebo n = 21). Since some of the Z scores in the left map are faint, the right map shows which Z scores are ≥3. The Roman numerals at the bottom of each lane denote antibodies against SARS-CoV-2 protein regions as listed in Figure S1. The maximum Z score in the left heatmap was 126. (B) Cumulative incidence of confirmed COVID-19 defined by symptomatic subjects testing positive solely by PCR. The BCG group (n = 96) had no symptomatic subjects who were PCR positive (0%), whereas the placebo group (n = 48) had 5 symptomatic and PCR-positive subjects (10.4%). This difference was significant (Fisher’s exact p = 0.0036) and translated to a vaccine efficacy of 100% with 0.99 posterior probability. There was low availability of positive PCR tests at point-of-care locations during the first 7 months of the trial. This, together with the need to perform the PCR test in a narrow window of about 2 weeks to be positive, is the reason why we designed the 9 non-PCR methods shown in Figures 2A, 2B, and 3A.

Figure 4

Cumulative infections and infection severity (A) Cumulative total infectious diseases during the 15 months of surveillance. This cumulative figure shows all infections per patient, including all COVID-19 events, within the BCG group (blue) compared with the placebo group (red). Included are the infections for which multiple subject events were documented in both BCG and placebo groups. Comparison by means of a Poisson model yields a significant difference with p = 0.004 (BCG n = 24 out of 96 T1D total, placebo n = 20 out of 48 T1D total). (B) Cumulative infections for two different time periods: pre-COVID-19 pandemic (the 2.5-year period prior to this trial, i.e., the pretrial) and during the COVID-19 pandemic (the current trial period of 15 months). The pretrial time period was when all clinical trial subjects received their ≥3 BCG vaccines or placebo vaccines; the current trial was when the subjects were monitored during the 15-month observation during the COVID-19 pandemic (total patients BCG n = 96, placebo n = 48). The lack of a statistical difference in the number of infections between BCG and placebo groups in the pretrial period suggests that a longer length of time than 2.5 years is necessary to realize BCG’s maximal infectious disease protection. It appears that during the entire period of 15 months, prior BCG vaccinations were protecting from COVID-19 and other infections. ^∗∗p < 0.01. (C) Infectious disease index for symptomatic patients. Symptomatic patients in the BCG-treated group had significantly reduced total infectious disease symptom index (placebo 152 ± 70 [n = 20] versus BCG 48 ± 11 [n = 31], p = 0.04, single tail and unpaired) as well as average infectious disease symptom index (placebo 23 ± 7 [n = 20] and BCG 13 ± 2 [n = 31], p = 0.04, single tail and unpaired). We first calculated total and average symptom scores per patient and then calculated average and SEM of each of these for BCG and placebo cohorts separately (^∗Student’s t test p < 0.05, one-tailed, unpaired). (D) Patients in the BCG cohort reported significantly fewer days of missed work during infections compared with the placebo group (^∗p = 0.02). Missed work was reported by 7 out of 32 BCG patients and by 8 out of 17 placebo patients. (E) The number of patients that reported at least one symptom were 20 out of 48 in the placebo group and 33 out of 96 in the BCG group. The placebo group had more severe average symptoms compared with the BCG group for 12 out of 12 symptoms (left panel). The number of patients in BCG and placebo groups that reported each symptom was then expressed (right panel). For 11 out of 12 symptoms, there was a higher percentage of symptomatic patients in the placebo group compared with the BCG group. Statistical analysis by Student’s t test (one-tailed, unpaired, ^∗p < 0.05).

Figure 5

Infection symptoms of trial participants versus adult household members (A) Infection symptoms in BCG and placebo groups compared with non-diabetic adult partners living in the same households. We collected surveys of symptoms of infectious diseases from all trial participants and household members of 13 BCG families and 7 placebo families. The BCG-treated trial participants had comparable or lower total infectious symptoms indexes compared with their partners living in the same household, whereas most placebo-treated trial participants had more severe symptoms compared with their partners. Statistical analysis of the differences by two-sample Wilcoxon testing was significant (two-tailed; p = 0.049; BCG n = 13 and placebo n = 7). Left panel: xy plot of symptom index of trial participants versus their household members. Right panel: the same data are also shown as an xy plot of household members versus the difference of trial participants and their household members. This plot makes it easier to visualize which groups had milder or more severe disease by introducing positive and negative differences. (B) Distribution of individual infectious disease symptoms in BCG- and placebo-treated participants and infected household members across the four scoring possibilities (0: no symptoms; 1: mild; 2: moderate; 3: severe symptoms). Number at risk data are shown in Figure S4.