Biometrics from a wearable device reveal temporary effects of COVID-19 vaccines on cardiovascular, respiratory, and sleep physiology

Abstract

Although vaccines against SARS-CoV-2 have been proven safe and effective, transient side-effects lasting 24-48 h postvaccination have been reported. To better understand the subjective and objective response to COVID-19 vaccination, we conducted a retrospective analysis on 69,619 subscribers to a wrist-worn biometric device (WHOOP Inc., Boston, MA) who received either the AstraZeneca, Janssen/Johnson & Johnson, Moderna, or Pfizer/BioNTech vaccine. The WHOOP device measures resting heart rate (RHR), heart rate variability (HRV), respiratory rate (RR), and sleep architecture, and these physiological measures were normalized to the same day of the week, 1 wk before vaccination. Averaging across vaccines, RHR, RR, and percent sleep derived from light sleep were elevated on the first night following vaccination and returned to baseline within 4 nights postvaccination. When statistical differences were observed between doses on the first night postvaccination, larger deviations in physiological measures were observed following the first dose of AstraZeneca and the second dose of Moderna and Pfizer/BioNTech. When statistical differences were observed between age groups or gender on the first night postvaccination, larger deviations in physiological measures were observed in younger populations and in females (compared with males). When combining self-reported symptoms (fatigue, muscle aches, headache, chills, or fever) with the objectively measured physiological parameters, we found that self-reporting fever or chills had the strongest association with deviations in physiological measures following vaccination. In summary, these results suggest that COVID-19 vaccines temporarily affect cardiovascular, respiratory, and sleep physiology and that dose, gender, and age affect the physiological response to vaccination.

NEW & NOTEWORTHY Here we report the first large-scale study investigating the effect of COVID-19 vaccines on cardiovascular, respiratory, and sleep physiology. We find that vaccines temporarily impact measures of cardiovascular, respiratory, and sleep physiology and that the degree of change in physiology is influenced by the manufacturer and dose of the vaccine and the gender and age of the vaccine recipient. These results provide insights into physiological changes that occur with COVID-19 vaccination and indicate that the unique responses that occur postvaccination may depend on manufacturer, dose, gender, and age.

INTRODUCTION

Vaccines are needed to curb the spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), which leads to coronavirus disease 2019 (COVID-19). Currently, vaccines from AstraZeneca, Janssen/Johnson & Johnson (J/J&J), Moderna, and Pfizer/BioNTech have been approved for use across Europe and in North and South America (1). While these vaccines effectively reduce SARS-CoV-2 infection (2–5), initial reports on reactogenicity indicate that local and systemic events can occur transiently upon vaccination (6). Previous data on reactogenicity rely on self-reported, subjective data and the effects of COVID-19 vaccine on objective physiological measures are unknown. As subjective data is prone to bias (7), there is a need for objective data collection that can provide insights into additional dimensions, such as sleep architecture, which individuals are not consciously aware of.

Wearables that continuously collect high-quality physiological data provide an opportunity to examine the effect of COVID-19 vaccination on cardiovascular, respiratory, and sleep measures at larger scales than would typically be possible in traditional clinical settings. Here we utilize objective data collected using the WHOOP Strap (WHOOP, Inc., Boston, MA), a validated (8), wrist-worn biometric device to determine the physiological effects of COVID-19 vaccination. This study is the first large-scale study to quantify the impact of available COVID-19 vaccinations using objective cardiovascular and sleep outcome data.

METHODS

Data Collection

Subjective responses to COVID-19 vaccines were collected via daily surveys delivered through the WHOOP mobile application to all WHOOP members (Boston, MA). Participants consented to their anonymized data being used for research purposes. Since data were not identifiable and were stored on a secure server, this study was deemed exempt from Institutional Review Board (IRB) oversight by Advarra’s IRB (Columbia, MD). Participants indicated that they had received a vaccine via a toggle on their app. Once a participant indicated they had received a vaccine, they were further asked to provide the date of the vaccination, the manufacturer of their vaccine, and if they experienced any symptoms following vaccination. The questions related to symptoms allowed for multiple selections and queried for experience of a fever, chills, muscle aches, headaches, or fatigue. Participants were also able to report no adverse events. Surveys were provided on a one-time basis for each dose when the user indicated they were vaccinated. A wrist-worn device (WHOOP strap 3.0; Boston, MA) that continuously collects heart rate, three-axis accelerometer, temperature, and three-axis gyroscope data was used to calculate cardiovascular, respiratory, and sleep measures (total sleep, light sleep, and restorative sleep, which is the summation of slow wave and rapid eye movement sleep). Resting heart rate is calculated as the mean value of heartbeats per minute during the last 5 min of the last episode of slow wave sleep. Heart rate variability is calculated as the root-mean-square of successive differences sampled during the last 5 min of the last episode of slow wave sleep. Respiratory rate is calculated as the median of respirations per minute calculated over the duration of sleep. The cardiovascular, respiratory, and sleep measures have been validated against gold standard polysomnography measures and have been found to have a low degree of bias and low precision errors (e.g., <20 min bias and precision errors for sleep duration) (8, 9).

Physiological metrics leading into and following vaccination were collated with the self-reported vaccine and symptomatology data. Baseline metrics to calculate percent changes were taken from the same day of the week, 1 wk before the vaccination. The same day of week was used to establish percent changes to control for confounding related to weekday/weekend differences in sleep behavior and outcomes, which have been previously reported in WHOOP members (10). Inclusion criteria consisted of being over the age of 18, having a WHOOP membership at the time of solicitation, and self-reporting having received one of the four vaccines included in this analysis. If a group within dose, age, gender, and manufacturer had a sample size <30 it was excluded from the analysis, which led to the removal of the AstraZeneca second dose females aged 18–29 and 55 and older from the analysis. Gender was a self-reported variable that allowed for participants to select male, female, nonbinary, or prefer not to choose, and, due to low number of responders (n < 30) that selected nonbinary and prefer not to choose, these groups were excluded from the analysis. Because of statistics being performed on percent change in a variable as compared with the same day of the week from the week prior, a value was not included in the analysis if the participant did not have a recording for the baseline measure (same day of the week, 1 week prior) or the night postvaccination being analyzed. Descriptive statistics (mean, sample size, and 95% confidence intervals) for each variable can be found in Supplemental Data File S1 (all Supplemental material is available at https://doi.org/10.6084/m9.figshare.16959058).

Statistical Analysis

All data were analyzed using the R programming language (version 4.0.4). Mixed-effects models were performed using the lme4 package (version 1.1–27.1) and estimated marginal means were calculated using the emmeans package (version 1.7.0). Age was discretized for all analyses by binning into 18–29, 30–39, 40–54, and 55 and older age groups. Confidence intervals (CIs) of 95% were used to report the error around the mean. Significance was set a P < 0.05 and was adjusted using Tukey’s method when performing multiple pairwise comparisons.

Subjective symptoms analysis.

To compare the proportion of users that experienced symptoms after being vaccinated, odds ratios were calculated from a logistic mixed-effects regression model that was designed using the binary response of whether a member experienced symptoms as the dependent variable; manufacturer and dose, discretized age, and gender as fixed independent variables; and random intercepts for each user. Likelihood ratio tests were used to determine the effects of each independent variable.

Magnitude and duration of physiological effects.

To determine the effect of the vaccines on physiological parameters, estimated marginal means and their confidence intervals were calculated from linear mixed-effects regression models that were designed with the percent change in cardiovascular, respiratory, or sleep measures as the dependent variable; discretized age, gender, dose, and manufacturer as fixed independent variables; and random intercepts for each user. ANOVA using Satterthwaite’s method for approximating degrees of freedom (11) was used to assess the effects of the independent variables on the percent deviation in physiological parameters on the first night postvaccination. Estimated marginal means were calculated by adjusting for or stratifying by duration, manufacturer, dose, gender, or age, and post hoc comparisons of estimated marginal means were performed using Tukey’s method to correct for multiplicity. Estimated marginal means and post hoc tests were used to determine the effects vaccines might have over time and within manufacturer, dose, gender, and age:

1. To assess the duration of an effect a vaccine may have, overall estimated marginal means were calculated for the 3-4 nights following vaccination by adjusting for discretized age, gender, dose, and manufacturer. Confidence intervals (CIs) were used to assess if a physiological parameter significantly deviated from zero (indicating an overall effect of vaccination) on the first night postvaccination. Marginal means and their 95% CIs were reported as is and were not compared between nights. When vaccines between manufacturers demonstrated opposite responses on the first night of vaccination, marginal means were not reported. In other words, if a physiological parameter deviated in a positive direction for one manufacturer and a negative direction for a different manufacturer, we do not report the overall effect of vaccination over multiple nights.

2. To assess the response to vaccination for each manufacturer, estimated marginal means stratified by dose were calculated for the first night postvaccination by adjusting for discretized age and gender. For this assessment, marginal means and their 95% CIs were reported as is and were not compared.

3. To assess differences in the response to vaccination between doses, estimated marginal means for each dose stratified by manufacturer were calculated for the first night postvaccination by adjusting for discretized age and gender.

4. To assess differences in the response to vaccination between gender, estimated marginal means for each gender stratified by manufacturer, and dose were calculated by adjusting for discretized age and gender.

5. To assess differences in the response to vaccination between age groups, estimated marginal means for each age group stratified within manufacturer and dose were calculated by adjusting for gender. Comparisons were made between the 18- to 29-year-old and 55 and older groups.

Relationship between subjective and objective symptoms.

To assess the relationship between subjective and objective responses to vaccination, coefficients were extracted from linear mixed-effects regression models that were designed with: the percent change in cardiovascular, respiratory, or sleep measures as the dependent variable; the subjective symptoms (fever, fatigue, chills, headache, and muscle ache) as fixed binary independent variables; age, gender, manufacturer, and dose as covariates; and random intercepts for each user. Significance of associations between physiological parameters and subjective symptoms were calculated via t tests based on Satterthwaite’s method (11). A symptom was deemed to associate with cardiovascular, respiratory, or sleep measures to a larger or smaller degree than another variable if the upper and lower limits of the 95% CIs for the coefficients did not overlap.

RESULTS

Self-Reported Adverse Events Postvaccination

In total, we received 99,435 responses (Table 1) to the voluntary COVID-19 vaccination survey and found significant differences in the odds of reporting an adverse event depending on manufacturer, dose, age, and gender. Compared with the second dose of Moderna, those who received AstraZeneca first dose [odds ratio (OR) = 1.22, 95% CI = 1.10–1.36; P < 0.001], AstraZeneca second dose (OR = 4.83, 95% CI = 3.84–6.06; P < 0.001), Pfizer/BioNTech first dose (OR = 7.70, 95% CI = 7.30–8.10; P < 0.001), Pfizer/BioNTech second dose (OR = 2.80, 95% CI = 2.62–2.91; P < 0.001), Moderna first dose (OR = 4.89, 95% CI = 4.62–5.17; P < 0.001), or J/J&J (OR = 1.70, 95% CI = 1.55–1.85; P < 0.001) vaccines were more likely to report being asymptomatic (Fig. 1). Men (OR = 1.46, 95% CI = 1.42–1.51; P < 0.001) were more likely than women to report being asymptomatic (Fig. 1). Compared with the 18- to 29-yr-old age group, those between the ages of 30 and 39 (OR= 1.18, 95% CI = 1.14 and 1.23; P < 0.001), 40–54 (OR = 1.50, CI = 1.44–1.55; P < 0.001), or 55 and older (OR = 2.36, CI = 2.22–2.50; P < 0.001) were more likely to report being asymptomatic (Fig. 1). Percentage of respondents that self-reported that they experienced fatigue, muscle aches, headaches, chills, or a fever postvaccination can be found in Supplemental Fig. S1.

Table 1.

Number of participants that responded to the survey questions for each manufacturer, dose, gender, and age

	AstraZeneca		Janssen/Johnson & Johnson	Moderna		Pfizer/BioNTech
Gender/Age	First Dose	Second Dose	Single Dose	First Dose	Second Dose	First Dose	Second Dose	Row Totals
Female
18–29	128	N/A	493	1,877	1,794	3,172	2,940	10,404
	3.70%		10.75%	10.65%	10.56%	10.8%	10.86%	10.52%
30–39	258	51	530	2,380	2,360	3,990	3,835	13,404
	7.46%	15.69%	11.56%	13.50%	13.89%	13.59%	14.16%	13.22%
40–54	421	35	328	1,480	1,469	2,403	2,347	8,483
	12.18%	10.77%	7.16%	8.39%	8.65%	8.18%	8.67%	8.47%
55 and older	63	N/A	65	326	402	462	492	1810
	1.82%		1.42%	1.85%	2.37%	1.57%	1.82%	1.82%
Male
18–29	331	46	1,089	3,223	2,749	5,605	4,660	1,7703
	9.57%	14.15%	23.76%	18.28%	16.18%	19.09%	17.21%	18.35%
30–39	680	78	1,209	4,624	4,398	7,598	6,937	25,524
	19.67%	24.00%	26.37%	26.23%	25.89%	25.87%	25.61%	25.44%
40–54	1,335	81	722	2,916	2,952	4,990	4,685	1,7681
	38.62%	24.92%	15.75%	16.54%	17.38%	16.99%	17.30%	17.93%
55 and older	241	34	148	806	863	1,146	1,188	4,426
	6.97%	10.46%	3.23%	4.57%	5.08%	3.90%	4.39%	4.48%
Total	3,457	325	4,584	17,632	16,987	29,366	27,084	99,435
	3.48%	0.33%	4.61%	17.73%	17.08%	29.53%	27.24%	100.00%

Percent annotation represents the percentage of column total. N/A, not available due to the low number of respondents (<30) within this group.

Figure 1.

Percentage of respondents self-reporting a lack of symptoms following vaccination. Data reflect percentage of responders within each cell (i.e., sex, age group, and dose). N/A, not available due to the low number of responses (n < 30) within this group. P values represent results from likelihood ratio tests.

Effect of COVID-19 Vaccination on Cardiovascular Parameters

On average across all vaccines, resting heart rate (RHR) was elevated on the first night postvaccination (estimated marginal mean = 8.8 ± 0.3%), returned to near baseline levels by the second night postvaccination (estimated marginal mean = 1.5 ± 0.3%), and remained near baseline levels on the third night postvaccination (estimated marginal mean = −0.8 ± 0.2%). Results from ANOVAs indicated that, on the first night postvaccination, the percent change in RHR depended on manufacturer [F_(3,63302) = 1377.28, P < 0.001], dose [F_(1,43722) = 5602.21, P < 0.001], age [F_(3,49339) = 308.0, P < 0.001], and gender [F_(3,48762) = 23.50, P < 0.001]. On the first night postvaccination, averaged across age and gender, RHR: increased 15.2 ± 0.7% after the single dose of J/J&J; increased 15.1 ± 0.8% after the second dose of Moderna; increased 14.2 ± 0.3% after the first dose of AstraZeneca; increased 7.2 ± 0.3% after the second dose of Pfizer/BioNTech; increased 3.9 ± 1.9% after the second dose of AstraZeneca; increased 3.4 ± 0.3% after the first dose of Moderna; and increased 1.6 ± 0.3% after the first dose of Pfizer/BioNTech. Comparing RHR response between doses within manufacturers, larger increases were observed after the first dose of AstraZeneca (P < 0.001) and the second dose of Moderna and Pfizer/BioNTech (both P < 0.001). Comparing between genders on the first night, females exhibited larger increases in their RHR after receiving either the second dose of Moderna (P = 0.001) or Pfizer/BioNTech (P < 0.001). Comparing between 18- to 29-yr-olds and those that are 55 and older, the younger group experienced larger increases in RHR after receiving the single dose of J/J&J (P < 0.001); the first dose of AstraZeneca (P < 0.001), Moderna (P < 0.001), and Pfizer/BioNTech (P = 0.01); and the second dose of Moderna (P < 0.001). The absolute and relative changes in RHR for all groups are illustrated in Fig. 2, A and B. In all, these results suggest that RHR increases the first night after vaccination and returns to near baseline levels within 3 nights, and that differences in the RHR response to vaccination may depend on manufacturer, dose, gender, and age.

Figure 2.

Cardiovascular measures following vaccination. A: resting heart rate (RHR) in the week leading up to and after vaccination. B: percent change in RHR as compared with the same day of the week, 1 wk prior. C: heart rate variability (HRV) in the week leading up to and after vaccination. D: percent change in HRV as compared with the same day of the week, 1 wk prior. Data are presented as means ± 95% confidence intervals. P values represent results from statistical tests examining effects of vaccination across and within subgroups on the first night postvaccination. P values for overall effects of vaccination on cardiovascular measures on the first night postvaccination were determined using the confidence intervals from estimated marginal means averaged over manufacturer, dose, gender, and age. P values for the effects of manufacturer, dose, gender, or age on the percent change in cardiovascular measures on the first night postvaccination were determined from ANOVAs [resting heart rate: manufacturer F_(3,63302) = 1377.28, dose F_(1,43722) = 5602.21, age F_(3,49339) = 308.0, and gender F_(3,48762) = 23.50; heart rate variability: manufacturer F_(3,63403) = 396.27, dose F_(1,42999) = 1562.89, age F_(3,49731) = 71.66, and gender F_(1,49149) = 44.25].

Because of the opposite responses observed between manufacturers in the percent change in heart rate variability (HRV) on the first night after vaccination (Fig. 2, C and D), estimated marginal means across manufacturers are not reported. Results from ANOVAs indicated that, on the first night postvaccination, the percent change in HRV depended on manufacturer [F_(3,63403) = 396.27, P < 0.001], dose [F_(1,42999) = 1562.89, P < 0.001], age [F_(3,49731) = 71.66, P < 0.001], and gender [F_(1,49149) = 44.25, P < 0.001]. On the first night postvaccination, averaged across gender and age, HRV: decreased 18.9 ± 2.4% after the first dose of AstraZeneca; decreased 16.2 ± 2.2% after the single dose of J/J&J; decreased 15.1 ± 1.0% after the second dose of Moderna; decreased 5.2 ± 6.2% after the second dose of AstraZeneca; decreased 4.2 ± 0.9% after the second dose of Pfizer/BioNTech; increased 1.9 ± 1.0% after the first dose of Moderna; and increased 5.0 ± 0.8% after the first dose of Pfizer/BioNTech. Comparing HRV responses between doses and within manufacturers, larger decreases were observed after the first dose of AstraZeneca (P < 0.001) and second dose of Moderna and Pfizer/BioNTech (both P < 0.001). On the first night postvaccination, differences in the percent change in HRV were observed between genders following the first dose of Moderna (P = 0.002) and second dose of Moderna and Pfizer/BioNTech (both P < 0.001). Comparing between 18- to 29-yr-olds and those that are 55 and older, the younger group experienced larger decreases in HRV after receiving the single dose of J/J&J (P < 0.001); the first dose of AstraZeneca (P = 0.02) and Moderna (P < 0.001); and the second dose of Moderna and Pfizer (both P < 0.001). The absolute and relative changes in HRV for all groups are illustrated in Fig. 2, C and D. In all, these results suggest that changes in HRV following vaccination depend on manufacturer, dose, gender, and age.

Effect of vaccines on respiratory rate.

On average across all vaccines, respiratory rate (RR) was increased on the first night postvaccination (estimated marginal mean = 3.5 ± 0.1%), remained elevated on the second night postvaccination (estimated marginal mean = 3.2 ± 0.1%), and returned to near baseline levels by the third (estimated marginal mean = 1.0 ± 0.1%) and fourth (estimated marginal mean = 0.3 ± 0.1%) night postvaccination. Results from ANOVAs indicated that, on the first night postvaccination, the percent change in RR depended on manufacturer [F_(3,62608) = 980.10, P < 0.001], dose [F_(1,41444) = 4534.69, P < 0.001], age [F_(3,48691) = 222.43, P < 0.001], and gender [F_(1,48105) = 17.59, P < 0.001]. On the first night postvaccination, averaged across gender and age, RR: increased 5.8 ± 0.3% after the single dose J/J&J; increased 5.7 ± 0.1% after the second dose of Moderna; increased 5.3 ± 0.3% after the first dose of AstraZeneca; increased 3.0 ± 0.1% after the second dose of Pfizer/BioNTech; increased 2.4 ± 0.9% after the second dose of AstraZeneca; increased 1.6 ± 0.1% after the first dose of Moderna; and increased 0.7 ± 0.1% after the first dose of Pfizer/BioNTech. Comparing RR responses between doses within manufacturers, larger increases were observed after the second dose of Moderna and Pfizer/BioNTech (both P < 0.001) and the first dose of AstraZeneca (P < 0.001). On the first night postvaccination, females exhibited larger percent increases in RR than males following the second dose of Moderna and Pfizer/BioNTech (both P < 0.001). Comparing between 18- to 29-yr-olds and those that are 55 and older, the younger group experienced larger increases in RR after receiving the single dose of J/J&J (P < 0.001); the first dose of AstraZeneca (P = 0.02), Moderna (P < 0.001), and Pfizer/BioNTech (P = 0.007); and the second dose of Moderna (P < 0.001) and Pfizer/BioNTech (P < 0.001). The absolute and relative changes in RR for all groups are illustrated in Fig. 3, A and B. In all, these results suggest that RR increases following vaccination and returns to near baseline levels within 4 nights postvaccination, and that differences in the RR response to vaccination may depend on manufacturer, dose, gender, and age.

Figure 3.

Respiratory rates following vaccination. A: respiratory rate (RR) in the week leading up to and after vaccination. B: percent change in RR as compared with the same day of the week, 1 wk prior. Data are presented as means ± 95% confidence intervals. P values represent results from statistical tests examining effects of vaccination across or within subgroups on the first night postvaccination. P value for overall effect of vaccination on respiratory rate on the first night postvaccination was determined using the confidence intervals from the estimated marginal mean averaged over manufacturer, dose, gender, and age. P values for the effects of manufacturer, dose, gender, or age on the percent change in respiratory rate on the first night postvaccination were determined from ANOVAs [manufacturer F_(3,62608) = 980.10, dose F_(1,41444) = 4534.69, age F_(3,48691) = 222.43, and gender F_(1,48105) = 17.59].

Effect of COVID-19 Vaccination on Sleep

Because of the opposite responses observed between manufacturers in the percent change in sleep duration on the first night after vaccination (Supplemental Fig. S2, A and B), estimated marginal means across manufacturers are not reported. Results from ANOVAs indicated that, on the first night postvaccination, the percent change in sleep duration depended on manufacturer [F_(3,64005) = 48.79, P < 0.001] and dose [F_(1,44713) = 4534.69, P < 0.001] but did not depend on age [F_(3,50448) = 1.40, P = 0.23] and gender [F_(1,49878) = 0.15, P = 0.63]. On the first night postvaccination, averaged across age and gender, sleep duration: increased 5.2 ± 1.0% after the first dose of Moderna; increased 5.0 ± 0.8% after the first dose of Pfizer/BioNTech; increased 3.8 ± 0.8% after the second dose of Pfizer/BioNTech; increased 1.3 ± 0.9% after the second dose of Moderna; decreased 2.7 ± 2.2% after the first dose of AstraZeneca; and was likely unchanged after the second dose of AstraZeneca (estimated marginal mean = 1.1 ± 5.7%) and single dose of J/J&J (estimated marginal mean = -1.4 ± 2.0%). The absolute and relative changes in sleep duration for all groups are illustrated in Supplemental Fig. S2, A and B.

On average across all vaccines, the percent change in the percentage of sleep derived from light sleep was elevated on the first night postvaccination (estimated marginal mean = 8.5 ± 0.6%), remained elevated on the second night postvaccination (estimated marginal mean = 5.3 ± 0.1%), and returned to near baseline levels by the third night postvaccination (estimated marginal mean = 2.4 ± 0.5%). Results from the ANOVAs indicated that, on the first night postvaccination, the percent change in the percentage of sleep derived from light sleep depended on manufacturer [F_(3,63518)=309.26, P < 0.001], dose [F_(1,44670) = 994.81, P < 0.001], and age [F_(3,49529) = 29.0, P < 0.001], but not gender [F_(1,48958) = 0.04, P = 0.85]. On the first night postvaccination, averaged across age and gender, the percent change in sleep derived from light sleep: increased 14.5 ± 1.4% after the single dose of J/J&J; increased 14.3 ± 1.5% after the second dose of AstraZeneca; increased 13.8 ± 0.6% after the first dose of Moderna; increased 6.7 ± 0.5% after the second dose of Pfizer/BioNTech; increased 4.1 ± 0.7% after the first dose of Moderna; increased 2.5 ± 0.5% after the first dose of Pfizer/BioNTech; and was likely unchanged after the second dose of AstraZeneca (estimated marginal mean = 2.3 ± 3.9%). Comparing the percent change in the percentage of sleep derived from light sleep between doses and within manufacturers, larger increases were observed after the first dose of AstraZeneca (P < 0.001) and second dose of Moderna and Pfizer/BioNTech (both P < 0.001). On the first night postvaccination, females exhibited larger percent increases percentage of sleep derived from light sleep than males following the second dose of Pfizer/BioNTech (P = 0.03). Comparing between 18- to 29-yr-olds and those that are 55 and older, the younger group experienced larger increases in the percentage of sleep derived from light sleep after receiving the single dose of J/J&J (P < 0.001) and the second dose of Moderna (P < 0.001) and Pfizer (P < 0.001). The absolute and relative changes in percent sleep derived from light sleep for all groups are illustrated in Fig. 4, A and B. In all, these results suggest that the percentage of sleep derived from light sleep increases on the first night following vaccination and returns to near baseline levels within 3 nights postvaccination, and that changes in the percentage of sleep derived from light sleep on the first night postvaccination may depend on manufacturer, dose, and age.

Figure 4.

Sleep architecture following vaccination. A: percent sleep derived from light sleep in the week leading up to and after vaccination. B: percent change in percent sleep derived from light sleep as compared with the same day of the week, 1 week earlier. C: percent sleep derived from restorative sleep in the week leading up to and after vaccination. D: percent change in percent sleep derived from restorative sleep as compared with the same day of the week, 1 week prior. Data are presented as means ± 95% confidence intervals. P values represent results from statistical tests examining effects of vaccination across and within subgroups on the first night postvaccination. P values for overall effects of vaccination on sleep architecture on the first night postvaccination were determined using the confidence intervals from estimated marginal means averaged over manufacturer, dose, gender, and age. P values for the effects of manufacturer, dose, gender, or age on the percent change in sleep architecture on the first night postvaccination were determined from ANOVAs [light sleep: manufacturer F_(3,63518) = 309.26, dose F_(1,44670) = 994.81, age F_(3,49529) = 29.0, and gender F_(1,48958) = 0.04; restorative sleep: manufacturer F_{(3, 67332)}=26.20, dose F_(1,52904) = 133.80, age F_(3,55929) = 6.74, and gender F_(1,55427) = 2.73].

Because of opposite effects observed between manufacturers for the percent change in percentage of sleep derived from restorative sleep (summation of slow wave and rapid eye movement sleep) on the first night postvaccination, estimated marginal means across vaccines are not reported. Results from the ANOVAs indicated that, on the first night postvaccination, the percent change in the percentage of sleep derived from restorative sleep depended on manufacturer [F_(3,67332) = 26.20, P < 0.001], dose [F_(1,52904) = 133.80, P < 0.001], and age [F_(3,55929) = 6.74, P < 0.001], but not gender [F_(1,55427) = 2.73, P = 0.10]. On the first night postvaccination, averaged across age and gender, the percent change in sleep derived from restorative sleep: decreased 8.2 ± 4.8% after the first dose of AstraZeneca; decreased 6.3 ± 1.94% after the second dose of Moderna; decreased 5.7 ± 4.4% after the single dose of J/J&J; increased 5.6 ± 1.7% after the first dose of Pfizer/BioNTech; increased 6.5 ± 2.0% after the first dose of Moderna; and was likely unchanged after the second dose of Pfizer/BioNTech (estimated marginal mean = 0.8 ± 1.7%) and second dose of AstraZeneca (estimated marginal mean = 3.6 ± 12.1%). The percent change in the percentage of sleep derived from restorative sleep following the first night of vaccination was different between the first and second dose of Moderna and Pfizer/BioNTech (P < 0.001). The absolute and relative changes in percent sleep derived from restorative sleep for all groups are illustrated in Fig. 4, C and D. In all, these results suggest that the percent change in percentage of sleep derived from restorative sleep following vaccination may depend on manufacturer, dose, and age.

Relationship between Subjective and Objective Measures

Since clinical trials often explicitly rely on survey data to determine reactogenicity, we wondered if subjective symptoms related to objective physiological measures. Interestingly, those who self-reported being asymptomatic still exhibited increases in their resting heart rate (β = 0.8 ± 0.3; P < 0.001) and percent sleep derived from light sleep (β = 1.1 ± 0.7; P = 0.001) (Fig. 5). Reporting fever or chills appeared to be associated with the largest deviations in RHR (chills: β = 4.8 ± 0.3; fever: β = 4.8 ± 0.3), HRV (chills: β = −7.3 ± 1.0; fever: β=-6.0 ± 1.1), RR (chills: β = 1.4 ± 0.1; fever: β = 1.2 ± 0.1), sleep duration (chills: β = −4.0 ± 1.0; fever: β = −4.6 ± 1.0), and percent of sleep from light sleep (chills: β = 4.4 ± 0.6; fever: β = 1.1 ± 0.7) or restorative sleep (chills: β = 1.1 ± 0.7; fever: β = 1.1 ± 0.7) (Fig. 5). In all, these results suggest that asymptomatic individuals may still experience deviations in their RHR and percent sleep from light sleep and that fever and chills have the strongest association with deviations in physiological measures.

Figure 5.

Relationship between objective and subjective measures. Estimates of the relationship between each symptom (none, fatigue, muscle ache, headache, fever, and chills) and each physiological measure (resting heart rate, heart rate variability, respiratory rate, sleep duration, percent sleep derived from light sleep, or percent sleep derived from restorative sleep). Points on plots represent the coefficients (estimates) from linear mixed-effect models that included each symptom (none, fatigue, muscle ache, headache, fever, and chills) as independent variables while accounting for manufacturer, dose, gender, and age. Error bars represent 95% confidence intervals. Statistics to assess if a physiological parameter and self-reported symptom associate with each other are calculated via t tests based on Satterthwaite’s approximation (11). Degrees of freedom for each test are as follows, resting heart rate: none = 77,832, fatigue = 77,821, muscle ache = 77,807, headache = 77,833, fever = 77,848, and chills = 77,850; heart rate variability: none = 77,758, fatigue = 77,851, muscle ache = 77,848, headache = 77,851, fever = 77,812, and chills = 77,817; respiratory rate: none = 77,579, fatigue = 77,750, muscle ache = 77,755, headache = 77,742, fever = 77,668, and chills = 77,674; sleep duration: none = 77,851, fatigue = 77,815, muscle ache = 77,799, headache = 77,829, fever = 77,859, and chills = 77,859; light sleep: none =77,852, fatigue = 77,716, muscle ache = 77,687, headache = 77,744, fever = 77,841, and chills = 77,834; restorative sleep: none = 77,273, fatigue = 77,099, muscle ache = 77,061, headache = 77,117, fever = 77,273, and chills = 77,252. *P < 0.05; **P < 0.01; ***P < 0.001.

DISCUSSION

The aim of this study was to examine the objective and subjective responses to COVID-19 vaccination in a large population of adults. To accomplish this, we performed a retrospective analysis of the acute effect of vaccination on self-reported, cardiovascular, respiratory, and sleep metrics on 69,619 individuals (providing 99,435 responses) receiving one of four COVID-19 vaccinations on or before May 17, 2021. Findings were divided into self-reported symptomatology and objectively measured sleep, respiratory, and cardiovascular metrics. The major findings of these analyses are that the physiological deviations observed following vaccination depended on the manufacturer and dose of the vaccine and the gender and age of the vaccine recipient. These analyses complement and extend the initial findings from the individual clinical trials (2–5) and self-reported surveys (6) by including more manufacturers and by examining objective measures. This is the first large-scale study to investigate the effect of COVID-19 vaccination on cardiovascular, respiratory, and sleep parameters and further research is required to determine the mechanisms behind these different responses.

When statistical differences were observed between genders, we found that females experience larger deviations in their physiological values postvaccination than males. Prior research demonstrates that females experience more adverse events than males following vaccination, possibly due to greater immune activation (12, 13). Our results indicate that the greater effects of vaccines on females than males are detectable in wrist-worn wearable data as well, although future research is warranted to determine whether these larger deviations in physiology coincide with greater immune responses to vaccination.

In line with prior reports on subjective measures, we find that older vaccine recipients were less likely to report adverse events than younger participants (14). We add to these reports by demonstrating that age also plays a role in the objectively measured physiological response to vaccination, whereby older individuals have an attenuated response as compared with younger participants when statistical differences between age groups were observed. Vaccine efficiency decreases with age (15), and it is plausible that the attenuated responses measured within this investigation are due to an impaired immune response postvaccination, but further research is required to test this hypothesis.

Another novel finding from this study is that certain subjective symptoms associate with the degree of the certain physiological responses. Additionally, asymptomatic individuals may still have discernible physiological responses to vaccination, indicating that self-reporting symptoms may not entirely reflect symptomology. Phase 3 clinical trials for vaccines rely on surveys due to the impracticality of serum profiling on large numbers of participants and the results from this manuscript indicate that information gathered from subjective surveying alone may underestimate the extent of side effects. Further research is required to understand the clinical significance of these observed physiological responses when not paired with perceived symptoms.

This is the first large-scale study to report on the acute objective and subjective reactogenicity of COVID-19 vaccination in adults. These results indicate that cardiovascular, respiratory, and sleep parameters deviate from baseline levels on the first night after COVID-19 vaccination, and that the degree of deviation is dependent on manufacturer, dose, gender, and age. Importantly, the perturbations to cardiovascular, respiratory, and sleep parameters were transient and returned to near prevaccine baseline levels within four nights following vaccination. These results are practically applicable to individuals anticipating receipt of a COVID-19 vaccine as they organize their postvaccine commitments as well as by policy makers and employers considering offering time off for vaccination.

Conclusions

Overall, we found that cardiovascular, respiratory, and sleep physiology were altered following COVID-19 vaccination. The magnitude of the changes in physiology depended on the manufacturer and dose of the vaccine and the gender and age of the participant. Larger deviations were observed for RHR, HRV, RR, and percent sleep from light sleep after the first dose of AstraZeneca and the second dose of Moderna and Pfizer (as compared with their respective first or second dose). When statistical differences were observed between genders, females experienced larger deviations in physiology as compared with males. Moreover, when statistical differences were observed between 18- to 29-yr-olds and those that are 55 and older, the younger group experienced larger deviations in physiology as compared to the older group. Importantly, deviations in physiology were short-lived and returned to near baseline levels within 4 nights postvaccination. These findings indicate that COVID-19 vaccines may have short-term effects that impact cardiovascular, respiratory, and sleep physiology.

Limitations

Interpretation of the results should be made with consideration to the limitations of this study. Extensive demographics were not collected for this study; therefore, we are unable to tell if certain demographics respond differently to vaccination. Data regarding the receipt of a COVID-19 vaccine, the preparation of vaccine received, and the symptoms experienced were collected via survey and not otherwise verified; however, we assume the frequency of recollection error was negligible due to the robustness of our findings and the short gap between vaccination and survey completion. As this study was retrospective and lacked a placebo control, we are unable to discern if the effects observed here are directly due to the vaccine or due to another factor (e.g., behavior modifications) that might influence cardiovascular, respiratory, or sleep measures. We utilize a single day from the week prior as a baseline, which may be more prone to random physiological fluctuations as compared with taking the same day of the week for multiple weeks leading up to vaccination; however, the use of a value closer to the time of vaccination would be less subject to long term physiological changes [e.g., increases in HRV due to exercise training (16)]. We describe many of the results using estimated marginal means averaged over multiple subgroups, which may lead to misinterpretations when interactions occur; therefore, when vaccines elicited opposite effects on an outcome of interest, we do not report the estimated marginal mean. To overcome this limitation, we report every pairwise comparison between age, gender, manufacturer, vaccine, and days from vaccination (not correcting for multiplicity) in Supplemental Data File S2. Additionally, although we adjust P values for multiple comparisons, we do not adjust for the confidence intervals around estimated marginal means, which may increase the amount of false positive results when interpreting the magnitude and direction of effects of vaccines.

DATA AVAILABILITY

Data and code can be made available upon reasonable request.

SUPPLEMENTAL DATA

Supplemental Data Files S1 and S2 and Supplemental Figs. S1 and S2: https://doi.org/10.6084/m9.figshare.16959058.

DISCLOSURES

All authors are affiliated with the commercial company WHOOP, Inc., which provided support in the form of salaries but did not otherwise play a role in the study design, data collection or analysis, decision to publish, or preparation of the manuscript.

AUTHOR CONTRIBUTIONS

D.M.P. and E.R.C. conceived and designed research; D.M.P. and E.R.C. analyzed data; D.M.P. and E.R.C. interpreted results of experiments; D.M.P. prepared figures; D.M.P. and E.R.C. drafted manuscript; D.M.P. and E.R.C. edited and revised manuscript; D.M.P. and E.R.C. approved final version of manuscript.