Study participants were recruited through public advertisements. Eligible individuals were men and women aged 18–55 years, non-smokers, free from illnesses, injuries, or disabilities affecting physical performance or caffeine metabolism, and with current experience of caffeine consumption (e.g., self-reported habitual intake of coffee, tea, energy drinks, or caffeine tablets). Individuals taking medications known to affect HR or interact with caffeine, or those with hypersensitivity to caffeine, were excluded. Additional contraindications were assessed via a health declaration.

All participants received written and verbal information about the study procedures, potential risks and discomforts, and their right to withdraw at any point. Written informed consent was obtained. The study adhered to the Declaration of Helsinki and was approved by the Regional Ethics Committee in Stockholm (reference number 2023-03730-01). Preregistration was completed on October 27, 2023, on the Open Science Framework (osf.io), DOI: https://doi.org/10.17605/OSF.IO/GU5YP.

This study employed a randomised crossover design in which all participants received each experimental treatment. Based on previous validation studies of the Ekblom–Bak (EB) test, the mean VO₂max in the relevant age group was 3.63 ± 1.05 L·min⁻¹, with a test–retest coefficient of variation of 6.2% (Björkman et al. 2016; Ekblom-Bak et al. 2014). The study was designed to detect a minimum change of 3% in VO₂max following caffeine intake or exercise. Using this information, the within-subject SD of differences was estimated as 0.225 L·min⁻¹. An a priori power analysis was performed using G*Power 3.1 (Heinrich Heine University, Düsseldorf, Germany) for a two-tailed paired t-test (α = 0.05, power = 0.80), which indicated that 37 participants were required to detect this effect size (Cohen’s d = 0.48).

The study was conducted in a double-blind and placebo-controlled manner with a crossover design. The EB-test was validated against measured VO2max, both with and without the influence of caffeine intake, and with and without prior physical exercise. The experimental conditions were caffeine (Caf) and placebo (Pla). Block randomisation was used to prevent severe imbalances in sample allocation. Each participant was randomly assigned to one of the ten equally sized, predetermined blocks. The random number list used to create these four blocks was created using an Excel spreadsheet with the RAND formula, “=RAND()”. A research assistant generated the random allocation sequence. An external researcher prepared identical administration tablets in numbered bags, labelling the two conditions as “A” and “B.” Participants were enrolled by the external researcher and assigned to interventions by a research assistant. Neither the participants nor the research assistant conducting the tests were aware of the allocation, ensuring a double-blind study.

In the weeks preceding the experiments, all participants visited the Laboratory for Applied Sports Science (LTIV) at the Swedish School of Sport and Health Sciences to familiarise themselves with the testing procedures. During this visit, participants completed an informed consent form, a health declaration, a cardiac screening questionnaire, and a brief questionnaire about caffeine habits. Subsequently, each participant performed an individually designed maximal running test to determine VO₂max (characterisation test).

The main experiments were conducted on two separate days, one week apart. Before the visit, participants were asked to refrain from caffeine, smoking, and vigorous physical activity for 24 h prior to the test, and not to consume any heavy meals within 3 h of the test. During the test day, participants were allowed to consume water ad libitum. All tests were performed in a climate-controlled laboratory environment, with the temperature maintained between 19.0 °C and 19.2 °C. The timeline of measurements for each experimental day is shown in Fig. 1. Body mass (measured while wearing lightweight clothes to the nearest 0.1 kg) and height (to the nearest 0.1 cm) were recorded, and participants were equipped with an HR monitor (Polar Electro, Kempele, Finland). All HR recordings were derived from the HR monitor. BP was measured on the right upper arm with an automatic oscillometric device (Omron M6 Comfort, model HEM-7221-E; Omron Corporation, Kyoto, Japan). The first measurements of HR and BP were conducted after 15 min of supine rest. Following 15 min of supine rest, baseline HR and BP were measured. Participants then ingested either 250 mg of caffeine (condition: Caf) or 50 µg of vitamin D (condition: Pla), administered in tablet form with a MEDCOAT® coating, which facilitates swallowing and masks flavour, and ensured that the tablets were visually identical in size and shape. After ingestion, participants rested for 45 min, after which HR and BP were reassessed.

Timeline of experimental procedures. Ex; exercise, EB-test; Ekblom-Bak test, HR; heart rate, BP; blood pressure, RPE; ratings of perceived exertion, VO2max; maximal oxygen uptake

Each experimental day included two submaximal EB-tests and one maximal treadmill running test to determine VO₂max. The EB-tests were conducted on a mechanically braked cycle ergometer (Monark Model 828E, Varberg, Sweden). After standard adjustment of the seat and handlebar, participants were provided with verbal instructions on the Borg’s RPE scale (Borg 1970) and were instructed to maintain a constant pedalling rate of 60 revolutions per minute throughout the test. The EB-test comprises 4 min of cycling at a standard work rate (0.5 kp, i.e., ~ 30 W), followed by 4 min of cycling at a higher, individually prescribed work rate to achieve a RPE of ≈ 14. The higher work rate was determined by the test leader based on sex, self-reported physical status, and exercise habits. HR was recorded as the average of the final minute at each work rate, and perceived exertion was rated using the RPE scale immediately after completing each work rate. Full details of the test procedure are described elsewhere (Ekblom-Bak et al. 2014). Participants cycled on a mechanically braked cycle ergometer (Monark Model 828E, Varberg, Sweden). The first cycle test (referred to as “EB pre-Ex`”) was performed 45 min after ingestion of caffeine or placebo. After a brief rest (maximum of 5 min), participants performed a strenuous exercise session (Ex) consisting of a graded maximal running test. The running test began with a 5-minute warm-up, during which participants started walking at a speed of 5–6 km/h, gradually increasing to a comfortable running pace. The test protocol was individually adjusted for starting speed and incline but followed a standardised format: the speed or incline was increased every minute until the participant reached voluntary exhaustion. Oxygen uptake was measured during the maximal running test using a Jaeger Oxycon Pro metabolic system (Hoechberg, Germany) with a Hans Rudolph facemask. Gas, volume, and ambient conditions were calibrated before each test. VO₂max was considered valid when a plateau in VO₂ occurred despite increasing speed or incline (Howley et al. 1995), supported by a respiratory exchange ratio ≥ 1.10, RPE > 16, exercise duration > 6 min, and a maximal HR within ± 10–15 beats per minute (bpm) of the age-predicted maximum (Poole and Jones 2017). The maximal test was followed by a 30-minute rest period, after which participants completed a second EB-test (“EB post-Ex”). VO₂max was estimated using validated sex-specific prediction equations that incorporate age and heart rate–work rate relationships derived from the two submaximal exercise stages (Björkman et al. 2016).

The study aimed to manipulate HR responses at two work rates during a submaximal cycle ergometer test by introducing 250 mg caffeine and/or strenuous exercise (Ex) prior to testing. The primary outcome variables in the statistical analyses were the estimation errors, i.e., the differences between the estimated and measured VO2max values under the different conditions. The secondary outcomes were resting cardiovascular variables (resting HR, systolic and diastolic BP), the submaximal HR values obtained during the EB test at low (30 W) and high work rates, as well as the exercise performance variables (time to exhaustion, measured VO₂max, and maximal HR) recorded during the experimental trials. Normality of the data was evaluated using the Shapiro–Wilk tests. Descriptive statistics are reported as mean ± standard deviation (SD) for continuous and normally distributed variables. Paired-samples t-tests were used to analyse differences in descriptive variables, with effect sizes reported as Cohen’s d. To account for multiple comparisons, p-values were adjusted using the Bonferroni–Holm correction method.

The multiple parameters from the EB-test and the differences in estimation error between conditions and time points were analysed using a two-way repeated-measures analysis of variance (RM ANOVA) with condition (Pla, Caf) and time (pre-Ex, post-Ex) as within-subject factors. Significant interaction effects were followed up with Tukey’s post hoc pairwise comparisons. In addition, equivalence testing was performed using the two one-sided tests (TOST) procedure. Equivalence bounds were set at ± 0.1 L·min⁻¹ to evaluate whether the estimation errors during Caf and Pla trials, in resting and exercised conditions, were statistically equivalent. For reference, previous validation studies report a typical estimation error of – 0.05 L·min⁻¹ among age-matched subjects (Björkman et al. 2016), and the ± 0.1 L·min⁻¹ bounds were applied to capture changes larger than the typical noise, but not trivially small fluctuations. To account for potential confounding effects of inter-individual differences, additional analyses were conducted using RM analysis of covariance (ANCOVA). Body mass and cardiorespiratory fitness level (i.e., measured VO2max) were entered as covariates to control for their potential influence on the estimation error. Although normality of the dependent variables was confirmed through Shapiro–Wilk tests, the normality of ANOVA residuals was also evaluated using Q–Q plots. Visual inspection indicated that residuals were approximately normally distributed. Effect sizes for ANOVA results are reported as partial eta squared (partial η²). Significance was set at p < 0.05. All statistical analyses were performed in Jamovi (version 2.6). This study was conducted in accordance with the CONSORT (Consolidated Standards of Reporting Trials) guidelines (Schulz et al. 2010).

Thirty-two habitual caffeine users volunteered to participate in the study (15 women and 17 men). Four subjects (two women and two men) had to withdraw from the study due to different illnesses and injuries. Of the 28 participants who completed the study, 13 received Caf first, and 15 received Pla first. In this crossover design, all randomised participants who completed all experimental sessions were included in analyses for all outcome variables. No adverse events or unintended effects were reported in any condition. The periods of participant recruitment and follow-up were from October 20, 2023, to March 14, 2024. The trial was completed as scheduled.

Subject characteristics are presented in Table 1. In this study, the administered dose of 250 mg caffeine corresponded to 3.33 ± 0.57 mg·kg⁻¹ body mass for the participants. The resting values of HR and BP are presented in Table 2.

No significant differences were observed for measured VO2max and maximal HR (3.81 ± 0.94 L min− 1 and 187 ± 13 bpm, p = 0.93, during the Pla trial and 3.81 ± 0.93 L min− 1 and 188 ± 12 bpm, p = 0.12, during the Caf trial, respectively). However, there was a significant difference in time to exhaustion: 415 ± 69 s during the Pla trial compared to 438 ± 73 s during the Caf trial (p = 0.012), respectively.

The primary outcome variables (measurement errors), as well as the submaximal HR recordings, Borg scale ratings and estimated VO2max values from the EB-test, are reported in Table 3. The unadjusted values from the RM ANOVA revealed a significant time effect, with post-exercise HR values increasing 13–27%, resulting in an approximately 5% greater underestimation of VO2max during the second EB-test, regardless of whether participants received caffeine or placebo. A significant time effect was also observed for RPE. HR at 30 W demonstrated an additional significant interaction effect. However, the post-hoc comparisons confirmed that HR increased significantly from pre- to post-Ex in both conditions, but there were no significant differences between Pla and Caf at either time point. Although the estimation error changed over time (mean difference: −0.11 L·min⁻¹ in the Pla condition and − 0.18 L·min⁻¹ in the Caf condition), the magnitude of change did not differ between conditions (Fig. 2). Equivalence testing using the TOST procedure revealed that in the pre-Ex Caf and Pla condition, the lower bound test was significant, t(27) = 2.82, p < 0.01, but the upper bound test was not significant, t(27) = -0.85, p = 0.20. When testing the differences in estimation error between post-Ex Caf and Pla conditions, the lower bound test was not significant (t(27) = 1.54, p = 0.067), while the upper bound test was significant (t(27) = -2.16, p = 0.02). Together, these results indicate that the observed differences in estimation errors fell outside the predefined equivalence bounds of ± 0.1 L·min⁻¹. The estimation errors were not affected by body mass or cardiorespiratory fitness (measured VO2max). After adjustment for these covariates, there were no significant main effects of condition (F(1, 25) = 0.001, p = 0.981) or time (F(1, 25) = 0.12, p = 0.733), and no significant interaction effect (F(1, 25) = 0.22, p = 0.645). Neither body mass nor measured VO₂max had a significant moderating influence on the effects of condition or time (all p > 0.14).

Estimated marginal means of estimation error across conditions (caffeine vs. placebo) and time points (pre- and post-exercise), derived from a repeated-measures ANOVA. Error bars represent 95% confidence intervals