Inter-individual variation in objective measure of reactogenicity following COVID-19 vaccination via smartwatches and fitness bands

The study addresses the need for scalable, non-invasive methods to objectively assess individual response (reactogenicity and potentially immunogenicity) to COVID-19 vaccination. The hypothesis is that digital biomarkers, especially deviations in resting heart rate (RHR) measured by consumer wearables relative to an individual’s baseline, can objectively capture physiologic responses to vaccination. Context includes widespread deployment of mRNA and adenoviral COVID-19 vaccines, variability in individual immune responses, and reliance on self-reported symptoms (e.g., CDC V-safe). The purpose is to determine whether wearable-derived RHR, sleep, and activity deviations around vaccination can serve as objective indicators of reactogenicity and to identify factors (prior infection, vaccine type, age, sex) associated with inter-individual variation.

Prior work shows considerable variability in immune responses to vaccination and mixed evidence linking reactogenicity symptoms to immunogenicity. CDC V-safe data reported high rates of systemic symptoms after mRNA vaccination, particularly after the second dose. A study using a smart ring found associations between physiologic changes and ~30-day antibody levels. Reactogenicity differences between Moderna and Pfizer-BioNTech have been reported, with higher subjective side effects for Moderna recipients. Literature also highlights mechanisms of vaccine-induced inflammation, including type I interferon responses and cytokine release, and the role of humoral and cellular immunity (neutralizing antibodies and T-cell responses). Fever is known to increase heart rate (~8.5 BPM per 1°C), but inflammation can elevate heart rate even without fever. Demographic factors such as age (immunosenescence) and sex differences have been observed in vaccine responses across vaccines.

Design and setting: Observational, app-based longitudinal prospective study (DETECT) in the U.S., enrolling 39,701 participants (March 25, 2020–September 12, 2021). Participants consented electronically; protocol approved by Scripps IRB (20-7531). Participants linked wearable devices and self-reported symptoms, COVID-19 test results, and vaccination dates/types via the DETECT app. Participants: 7,728 reported at least one COVID-19 vaccine dose; 7,298 received an mRNA vaccine (Pfizer-BioNTech or Moderna). After applying exclusion criteria for data completeness and quality, 5,674 individuals were included for RHR analysis; 4,628 for activity; 5,691 for sleep. Devices: Fitbit (76%) and Apple Watch (20%); 152 other devices excluded from this analysis. Due to small numbers, Johnson & Johnson recipients (n=437) were excluded from main comparisons but summarized in supplementary analyses. Inclusion/exclusion criteria for analytics: Excluded 75 with vaccine dates before Dec 11, 2020, and 16 without age or gender. Valid daily values required ≥15 hours wear-time. Individuals were excluded if they had: fewer than 4 recording days in the 2 weeks before vaccination; fewer than 3 of the 5 days after vaccination; or fewer than 14 days during the baseline period (from 60 to 7 days pre-vaccination). Missing data led to exclusion from specific metric analyses (RHR: 1,552; sleep: 2,598; activity: 1,535 individuals). Metrics and baseline computation: Daily metrics included RHR (Fitbit: morning supine estimate; Apple: proprietary daily estimate), total sleep minutes, and steps. For each participant, an individual baseline was computed over days −60 to −7 relative to vaccination using exponentially decayed weighting (exponent α=0.05), giving more weight to recent days. Daily deviations were computed as metric(d) − baseline. For summary analyses, the average of absolute changes over day 0 through day +4 (vaccination day and the following 4 days) was used. Subgrouping: Analyses were performed separately for first and second doses. Subgroups: sex (female/male), age (<40, 40–65, >65), vaccine type (Pfizer-BioNTech vs Moderna), and prior reported COVID-19 positive test (yes/no). Statistical analysis: For time series plots (−14 to +14 days relative to dose), mean and 95% CI of metric deviations were shown. For subgroup comparisons, means (with 95% CI via bootstrap, 10,000 iterations) of individual averages over day 0–4 were calculated. Two-sided t-tests assessed differences among groups; chi-squared tests assessed changes in subgroup frequencies. Multiple hypothesis testing was corrected by Holm–Bonferroni within three families (RHR, sleep, activity). Multiple linear regression estimated adjusted associations (estimated marginal means) controlling for age, sex, device, vaccine type, and prior infection. Device-related differences (e.g., Apple vs Fitbit) were assessed. Significance thresholds and adjusted p-values reported; some sex differences lost significance after correction.

Physiologic response (RHR):

• Population-level RHR increased the day after vaccination, peaking on day 2, and returned to baseline by day 4 (first dose) and day 6 (second dose).
• Peak mean increases on day 2: +0.56 BPM (95% CI 0.48–0.65; p<0.001) after first dose; +1.52 BPM (95% CI 1.42–1.63; p<0.001) after second dose.
• Proportion with RHR increase within 2 days post-dose: 71% (first dose) and 76% (second dose).
• Proportion exceeding >1 SD above individual’s normal daily RHR variation: 37% (first dose) and 47% (second dose). Subgroup differences (unadjusted unless noted):
• Prior COVID-19 infection: Higher average RHR increase after first dose (0.66 vs 0.28 BPM; p=0.008); no significant difference after second dose (0.43 vs 0.62 BPM; p=0.181). Adjusted regression: first dose estimated marginal mean 0.68 (95% CI 0.41–0.95) vs 0.30 (95% CI 0.22–0.37); p=0.007. Second dose 0.53 (95% CI 0.25–0.81) vs 0.72 (95% CI 0.64–0.80); p=0.188.
• Vaccine type: Moderna associated with greater RHR increases than Pfizer-BioNTech after both doses: first dose 0.41 vs 0.22 BPM (p=0.003); second dose 0.85 vs 0.44 BPM (p<0.001). Adjusted regression: first dose 0.58 (95% CI 0.42–0.75) vs 0.39 (95% CI 0.24–0.55); p=0.003. Second dose 0.84 (95% CI 0.67–1.01) vs 0.42 (95% CI 0.26–0.58); p<0.001.
• Sex: Women showed higher RHR increase after the first dose (p=0.014), but this became non-significant after Holm–Bonferroni correction. Adjusted regression first dose: 0.58 (95% CI 0.42–0.74) vs 0.43 (95% CI 0.26–0.60); p=0.021. No significant sex difference after second dose.
• Age: Individuals <40 years had the largest RHR increases after the second dose; difference vs ≥40 years ~+0.79 BPM (p=0.011). Adjusted regression second dose: 0.84 (95% CI 0.65–1.04) vs 0.60 (95% CI 0.45–0.76); p=0.005. No significant differences after first dose.
• Device: Apple devices showed higher average RHR changes than Fitbit, significant after second dose (p<0.001), likely due to differing RHR algorithms. Behavioral changes (sleep and activity):
• First dose: Minimal changes; day 1 mean sleep +8 minutes (95% CI 6–11); no decrease in steps.
• Second dose: Day 1 significant reductions in activity and increases in sleep that reverted by day 2: steps −1628 (95% CI −1726 to −1530); sleep +35 minutes (95% CI 33–39).
• Correlations between RHR changes and sleep/activity changes were low (Pearson r with sleep −0.05 first dose, −0.01 second; with activity 0.02 first dose, 0.01 second).
• Moderna recipients showed greater second-dose sleep increase and step decrease than Pfizer-BioNTech. Older adults (>65) had distinct step patterns (e.g., negative change after first dose). Some subgroup differences remained significant after multiple testing correction (e.g., vaccine-type effects on steps and sleep for second dose). Other vaccine: Limited J&J single-dose recipients showed RHR changes comparable to the second mRNA dose in supplementary analyses, aligning with reported subjective reactogenicity. Explained variance: Observable covariates explained a small fraction of variance in RHR responses (about 1.2% for average changes; <21.1% for peak changes), indicating substantial inter-individual variability.

Findings support the hypothesis that subtle deviations in wearable-derived RHR provide objective evidence of vaccine reactogenicity at the individual level. The temporal pattern (rapid increase peaking on day 2, normalizing within a week) and stronger effects after the second mRNA dose parallel known inflammatory responses and reported subjective symptoms. Greater responses among Moderna recipients and individuals with prior COVID-19 (after first dose) are consistent with higher antigen dose and primed immunity, respectively. Younger age and, to a lesser extent, female sex showed stronger responses, reflecting known immunological differences. While behavioral changes were evident mainly after the second dose, their weak correlation with RHR suggests RHR deviations capture physiological inflammation rather than behavior alone. The substantial unexplained inter-individual variability raises the possibility that RHR response magnitude may relate to immunogenicity; if validated against humoral and cellular measures, wearables could offer a scalable surrogate to identify suboptimal or exaggerated responses and guide personalized vaccination strategies (e.g., timing of boosters).

Consumer wearables can detect small but significant, individualized RHR deviations following COVID-19 vaccination that reflect objective reactogenicity. Responses are typically larger after the second dose, greater for Moderna than Pfizer-BioNTech, higher after the first dose among previously infected individuals, and more pronounced in younger adults. Sleep increases and activity decreases occur mainly after the second dose but correlate weakly with RHR changes. These results demonstrate feasibility of scalable, non-invasive monitoring of vaccine responses. Future research should validate these digital biomarkers against immunologic endpoints (neutralizing antibodies, T-cell responses), assess generalizability across devices and populations, evaluate predictive value for protection and breakthrough risk, and explore clinical applications such as identifying weak responders who may benefit from tailored booster timing or additional evaluation.

• Self-reported vaccination dates/types and prior infection status may be inaccurate; undiagnosed prior infections could bias first-dose responses.
• Study population comprises wearable device users, potentially limiting generalizability (lower use among older and lower socioeconomic groups).
• Only daily aggregated metrics were analyzed; intra-day data were not used, potentially missing finer-grained dynamics.
• No data on use of antipyretic or anti-inflammatory medications, which could attenuate physiological responses.
• Device heterogeneity and proprietary algorithms (Fitbit vs Apple Watch) may influence RHR estimates; device-related differences were observed.
• Johnson & Johnson recipients were too few for robust comparison; main analyses focused on mRNA vaccines.
• Observational design limits causal inference; unmeasured confounding may remain despite adjustments.