PPG-Based Heart Rate Accuracy in Hispanic Adults with Fitzpatrick III–V Skin Tones: An Evaluation of Body Composition and Skin-Tone Effects

Probably not a substantial claim: may not pass S3 (load-bearing).

A power analysis conducted prior to data collection indicated that a minimum of 52 participants would be needed to detect a medium effect size (f2(v) = 0.20) in a MANOVA with two groups and two dependent variables, assuming α = 0.05 and power (1 − β) = 0.80.

“A power analysis conducted prior to data collection indicated that a minimum of 52 participants would be needed to detect a medium effect size (f2(v) = 0.20) in a MANOVA with two groups and two dependent variables (absolute and directional error), assuming α = 0.05 and power (1 − β) = 0.80 [23].”

• cites [23] Bent, 2020, Investigating sources of inaccuracy in wearable optical heart rate sensors, NPJ Digit. Med. sourceDisputed

  Split into 2 atoms — 0 confirmed, 1 partially supported, 1 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Checked by a reasoning model — may be wrong.

  Experimental — accuracy not yet measured

May combine two claims.

Obesity alters dermal parameters including increased dermal thickness, greater trans-epidermal water loss, reduced capillary density, and increased depth of arteries such as the radial artery.

“Obesity alters several key dermal parameters: increased dermal thickness, greater trans-epidermal water loss (TEWL), reduced capillary density, and increased depth of arteries such as the radial artery.”

• cites [16] Ndumele, 2016, Obesity and subtypes of incident cardiovascular disease, J. Am. Heart Assoc. sourceCould not verify

  The Discussion addresses obesity-related cardiac remodeling and heart failure, not dermal changes. Searches for dermal thickness, trans-epidermal water loss, capillary density, and radial-artery depth found no passage in our copy addressing those claims. (LLM, conf=0.77)

  Experimental — accuracy not yet measured
• cites [17] Yosipovitch, 2007, Obesity and the skin: Skin physiology and skin manifestations of obesity, J. Am. Acad. Dermatol. sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Obesity is a known independent risk factor for multiple cardiovascular diseases.

“Obesity is a known independent risk factor for multiple cardiovascular diseases [16,21,22]”

• cites [16] Ndumele, 2016, Obesity and subtypes of incident cardiovascular disease, J. Am. Heart Assoc. sourcePartly supported

  The Results report that obesity remained associated with incident heart failure after adjustment for traditional cardiovascular risk factors. They found no significant adjusted association with coronary heart disease or stroke. The claim’s description of obesity as an established independent risk factor for multiple cardiovascular diseases therefore overstates this paper’s findings. (LLM, conf=0.88)

  Experimental — accuracy not yet measured
• cites [21] Wu, 2015, High blood pressure and all-cause and cardiovascular disease mortalities in community-dwelling older adults, Medicine sourceCould not verify

  The Introduction identifies high blood pressure—not obesity—as a cardiovascular disease risk factor. The Results analyze BMI as a subgroup in the relationship between hypertension and CVD mortality, but do not establish obesity as an independent risk factor for multiple cardiovascular diseases (searched: obesity, obese, BMI, independent risk). (LLM, conf=0.76)

  Experimental — accuracy not yet measured
• cites [22] Ogden, 2017, Prevalence of obesity among adults, by household income and education—United States, 2011–2014, MMWR Morb. Mortal. Wkly. Rep. sourceCould not verify

  The Discussion mentions possible health risks associated with BMI, but does not identify cardiovascular diseases or describe obesity as an independent risk factor. Searches for cardiovascular disease, heart disease, and stroke found no passage addressing that claim; the paper instead analyzes obesity prevalence by income and education. (LLM, conf=0.74)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing).

Age-related changes, such as decreased vascular compliance, increased arterial stiffness, and thinning of the epidermis, reduce the pulsatility and clarity of the PPG waveform.

“Age-related changes, such as decreased vascular compliance, increased arterial stiffness [18], and thinning of the epidermis [19], reduce the pulsatility and clarity of the PPG waveform.”

• cites [18] Vlachopoulos, C., O’Rourke, M., and Nichols, W.W. (2011). McDonald’s Blood Flow in Arteries: Theoretical, Experimental and Clinical Principles, CRC Press. sourceNot checked — no full text of the source
• cites [19] Derraik, J.G., Rademaker, M., Cutfield, W.S., Pinto, T.E., Tregurtha, S., Faherty, A., Peart, J.M., Drury, P.L., and Hofman, P.L. (2014). Effects of age, gender, BMI, and anatomical site on skin thickness in children and adults with diabetes. PLoS ONE, 9. sourcePartly supported

  The Results (Effects of age, puberty, and gender; page unavailable) report that dermal thickness decreases with age in adults. The Methods say the dermal measurement includes a small contribution from the epidermis, but the paper does not establish that the epidermis itself thins or that thinning reduces PPG waveform pulsatility or clarity. The vascular changes are attached to a different citation. (LLM, conf=0.86)

  Experimental — accuracy not yet measured

May combine two claims.

Pooled comparisons by BMI category did not reach significance; however, several device-level contrasts indicated small directional biases with higher adiposity.

“In contrast, pooled comparisons by BMI category did not reach significance; however, several device-level contrasts indicated small directional biases with higher adiposity (e.g., [12,13]).”

• cites [12] Rodriguez, 2021, Monte Carlo analysis of optical heart rate sensors in commercial wearables: The effect of skin tone and obesity on the photoplethysmography (PPG) signal, Biomed. Opt. Express sourceNot checked — no full text of the source
• cites [13] Rodriguez, 2021, Synthetic photoplethysmography (PPG) of the radial artery through parallelized Monte Carlo and its correlation to body mass index (BMI), Sci. Rep. sourcePartly supported

  The Results (Figures 6 and 8–9; page not provided) show that simulated PPG signals decline as modeled obesity increases, supporting the direction of the reported effect. The paper does not report pooled statistical comparisons by BMI category or device-level contrasts, and its modeled signal changes are not described as small device biases. (LLM, conf=0.84)

  Experimental — accuracy not yet measured

May combine two claims.

Statistically reliable patterns tied to phenotype, particularly adiposity, and to a lesser extent skin pigmentation, were evident and align with optical and physiological mechanisms described in prior work.

“Nonetheless, statistically reliable patterns tied to phenotype, particularly adiposity, and, to a lesser extent, skin pigmentation, were evident and align with the optical and physiological mechanisms described in prior work [13,43].”

• cites [13] Rodriguez, 2021, Synthetic photoplethysmography (PPG) of the radial artery through parallelized Monte Carlo and its correlation to body mass index (BMI), Sci. Rep. sourceSupported
• cites [43] Fine, J., Branan, K.L., Rodriguez, A.J., Boonya-Ananta, T., Ajmal Ramella-Roman, J.C., McShane, M.J., and Cote, G.L. (2021). Sources of inaccuracy in photoplethysmography for continuous cardiovascular monitoring. Biosensors, 11. sourceSupported

May combine two claims.

Apple and Garmin exhibited the lowest mean absolute error (MAE) among the four devices, whereas Fitbit and Samsung showed greater error.

“Apple and Garmin exhibited the lowest MAE, whereas Fitbit and Samsung showed greater error.”

May combine two claims.

In a linear mixed-effects model, Fitzpatrick V participants had approximately 1.6 percentage points higher MAPE than types III-IV (p ≈ 0.05), while BMI was not a significant predictor.

“The mixed-effects analysis indicated that Fitzpatrick V participants had approximately 1.6 percentage points higher MAPE than types III–IV (p ≈ 0.05) while BMI was not a significant predictor”

• AssumptionRests on an assumption: Self-reported Fitzpatrick skin type accurately classifies skin tone for the analyses.

May combine two claims.

Significant effects of BMI category were observed for ICC (U(n1 = 52, n2 = 88) = 3132, p < 0.001, rrb = −0.369), indicating that participants with higher BMI exhibited reduced measurement reliability.

“and ICC, U (n1 = 52, n2 = 88) = 3132, p < 0.001, rrb = −0.369”

• AssumptionRests on an assumption: BMI is a valid proxy for body composition/adiposity in this cohort.

May combine two claims.

Significant differences in bias were observed for the Fitbit device (U(n1 = 23, n2 = 13) = 81.0, p = 0.025, rrb = 0.458), with participants with BMI > 30 exhibiting greater directional error than participants with BMI < 25.

“and Fitbit, U(n1 = 23, n2 = 13) = 81.0, p = 0.025, rrb = 0.458”

• AssumptionRests on an assumption: BMI is a valid proxy for body composition/adiposity in this cohort.

May combine two claims. The quote only partly states this claim.

Significant differences in bias were observed for the Garmin device (U(n1 = 23, n2 = 13) = 63.0, p = 0.005, rrb = 0.579), with participants with BMI > 30 exhibiting greater directional error than participants with BMI < 25.

“Garmin, U(n1 = 23, n2 = 13) = 63.0, p = 0.005, rrb = 0.579”

• AssumptionRests on an assumption: BMI is a valid proxy for body composition/adiposity in this cohort.

May combine two claims. The quote only partly states this claim.

Significant effects of BMI category were observed for bias (U(n1 = 52, n2 = 88) = 1337, p < 0.001, rrb = 0.416), indicating that participants with higher BMI exhibited greater directional bias.

“Significant effects of BMI were observed for bias, U(n1 = 52, n2 = 88) = 1337, p < 0.001, rrb = 0.416”

• AssumptionRests on an assumption: BMI is a valid proxy for body composition/adiposity in this cohort.

May combine two claims. The quote only partly states this claim.

Significant differences in bias were observed for the Apple device (U(n1 = 22, n2 = 13) = 72.0, p = 0.016, rrb = 0.497), with participants with BMI > 30 exhibiting greater directional error than participants with BMI < 25.

“Significant differences in bias were observed for the Apple, U(n1 = 22, n2 = 13) = 72.0, p = 0.016, rrb = 0.497”

• AssumptionRests on an assumption: BMI is a valid proxy for body composition/adiposity in this cohort.

The quote only partly states this claim.

Across devices, participants with skin type III exhibited mean MAPE of 2.87% [2.31, 3.55].

“participants with skin types III and IV exhibited mean MAPE of 2.87% [2.31, 3.55]”

• AssumptionRests on an assumption: Self-reported Fitzpatrick skin type accurately classifies skin tone for the analyses.

May combine two claims.

For the Garmin device, heart rate was not a significant predictor of bias, but was significantly associated with both √AE and √APE, suggesting modest improvements in accuracy with higher heart rate.

“For Garmin device, HR was not a significant predictor of bias, but was significantly associated with both √AE and √APE, suggesting modest improvements in accuracy with higher HR.”

• AssumptionRests on an assumption: The K5 metabolic system heart rate accurately indexes exercise intensity.

May combine two claims.

In the study sample, bias increased and ICC declined as body fat percentage rose.

“The scatterplots in Figure 7 show that bias increased and ICC declined as BF% rose”

May combine two claims.

No significant gender differences were found for MAE, MAPE, or bias (U = 3825.5, p = 0.437, rrb = −0.062; U = 3942.5, p = 0.648, rrb = −0.035; U = 3675.5, p = 0.228, rrb = −0.092).

“No significant gender differences were found for MAE, U = 3825.5, p = 0.437, rrb= −0.062; MAPE, U = 3942.5, p = 0.648, rrb= −0.035, or bias, U = 3675.5, p = 0.228, rrb= −0.092.”

May combine two claims.

For the Samsung device, heart rate significantly predicted all three error metrics, indicating that directional, absolute, and relative errors decreased modestly with increasing heart rate.

“For the Samsung device, HR significantly predicted all three error metrics, indicating that directional, absolute, and relative errors decreased modestly with increasing HR.”

• AssumptionRests on an assumption: The K5 metabolic system heart rate accurately indexes exercise intensity.

May combine two claims.

Device-level gender comparisons did not yield any statistically significant differences in error metrics (ps > 0.05), although several comparisons showed small-to-moderate effect sizes.

“Device-level gender comparisons did not yield any statistically significant differences in error metrics (ps > 0.05), although several comparisons showed small-to-moderate effect sizes.”

May combine two claims.

For the Apple device, heart rate did not significantly predict bias, but was significantly associated with √AE; no significant effect was observed for √APE.

“For the Apple device, HR did not significantly predict bias, but was significantly associated with √AE; no significant effect was observed for √APE.”

• AssumptionRests on an assumption: The K5 metabolic system heart rate accurately indexes exercise intensity.

May combine two claims.

The hierarchical models had marginal R2 values ranging from 0.011 to 0.113 and conditional R2 values ranging from 0.07 to 0.26.

“Model fits indicated that marginal R2 values ranged from 0.011 to 0.113, reflecting the proportion of variance explained by the fixed effects alone (see Table 5). Conditional R2 values, which include both fixed and random effects, ranged from 0.07 to 0.26”

May combine two claims.

Gender had a significant effect on ICC (U(n1 = 96, n2 = 88) = 3104.5, p = 0.004, rrb = −0.230), indicating that female participants had significantly lower heart-rate reliability than male participants.

“A significant effect of gender was observed for ICC, U(n1 = 96, n2 = 88) = 3104.5, p = 0.004, rrb= −0.230, indicating that female participants had significantly lower HR reliability than male participants.”

May combine two claims.

For the Fitbit device, heart rate significantly predicted all three error metrics, with particularly strong reductions in relative error.

“For the Fitbit device, HR significantly predicted all three error metrics, with particularly strong reductions in relative error.”

• AssumptionRests on an assumption: The K5 metabolic system heart rate accurately indexes exercise intensity.

May combine two claims.

The present study found that all devices deviated from the ECG criterion, systematic bias was small in magnitude, and Bland-Altman dispersion was generally narrow relative to criterion heart rate.

“found that, although all devices deviated from the ECG criterion, systematic bias was small in magnitude and the Bland–Altman dispersion shown in Figure 4 and Figure 6 was generally narrow relative to criterion HR.”

• AssumptionRests on an assumption: The ECG chest-strap criterion provides valid heart rate measurement for agreement analysis.

May combine two claims.

Adiposity increases bias and decreases reliability in 3/4 devices.

“adiposity increases bias and decreases reliability in 3/4 devices”

May combine two claims.

Regression analyses indicated that relative error decreased with increasing intensity across devices, although these slopes were small (<0.01 units per bpm) and varied by device.

“Regression analyses indicated that the relative error decreased with increasing intensity across devices, although these slopes were small (<0.01 units per bpm) and varied by device.”

• AssumptionRests on an assumption: The K5 metabolic system heart rate accurately indexes exercise intensity.

May combine two claims.

For the Samsung device, all three metrics exhibited significant three-way interactions between exercise intensity, BF%, and Fitzpatrick skin tone: bias b = −0.60, t(4799) = −3.40, p < 0.001; √AE b = 0.09, t(4795) = 3.66, p < 0.001; and √APE b = 0.07, t(4796) = 3.01, p = 0.003.

“For the Samsung device, all three metrics exhibited significant three-way interactions between exercise intensity, BF%, and Fitzpatrick skin tone bias, b = −0.60, t(4799) = −3.40, p < 0.001; √(AE), b = 0.09, t(4795) = 3.66, p < 0.001; and √(APE), b = 0.07, t(4796) = 3.01, p = 0.003.”

May combine two claims.

The mixed-effects models explained little variance via fixed effects, with far larger conditional R2 driven by participant-level random effects, indicating substantial individual differences not captured by the measured covariates.

“Our mixed-effects models explained little variance via fixed effects, with far larger conditional R2 driven by participant-level random effects, indicating substantial individual differences not captured by the measured covariates.”

May combine two claims.

For the Apple device, significant three-way interactions between exercise intensity, BF%, and Fitzpatrick skin tone were observed for bias (b = 0.27, t(4947) = 2.69, p = 0.007) and √APE (b = −0.04, t(4968) = −2.22, p = 0.026); there was also a significant two-way interaction between Fitzpatrick and exercise intensity for √AE (b = −0.04, t(4963) = −2.96, p = 0.003).

“For the Apple device, significant three-way interactions between exercise intensity, BF%, and Fitzpatrick skin tone were observed for both bias, b = 0.27, t(4947) = 2.69, p = 0.007) and √(APE) (b = −0.04, t(4968) = −2.22, p = 0.026. There was also a significant two-way interaction between Fitzpatrick and exercise intensity for √(AE), b = −0.04, t(4963) = −2.96, p = 0.003.”

May combine two claims.

For the Fitbit device, all three metrics indicated significant two-way interactions of exercise intensity and Fitzpatrick skin tone: bias b = 0.73, t(4957) = 5.16, p < 0.001; √AE b = −0.05, t(4950) = −2.61, p = 0.009; √APE b = −0.07, t(4950) = −3.51, p < 0.001.

“For the Fitbit device, all three metrics bias, b = 0.73, t(4957) = 5.16, p < 0.001; √(AE), b = −0.05, t(4950) = −2.61, p = 0.009; √(APE): b = −0.07, t(4950) = −3.51, p < 0.001 indicated significant two-way interactions of exercise intensity and Fitzpatrick skin tone.”

May combine two claims.

For the Garmin device, all three metrics demonstrated significant three-way interactions between exercise intensity, BF%, and Fitzpatrick skin tone: bias b = 0.51, t(4935) = 4.46, p < 0.001; √AE b = −0.06, t(4962) = −3.07, p = 0.002; and √APE b = −0.07, t(4963) = −3.85, p < 0.001.

“For the Garmin device, all three metrics demonstrated significant three-way interactions between exercise intensity, BF%, and Fitzpatrick skin tone bias, b = 0.51, t(4935) = 4.46, p < 0.001; √(AE), b = −0.06, t(4962) = −3.07, p = 0.002; and √(APE), b = −0.07, t(4963) = −3.85, p < 0.001.”

May combine two claims.

When devices were pooled, agreement degraded with darker pigmentation within the Fitzpatrick III-V range: Bland-Altman dispersion was broader at Fitzpatrick V and systematic bias was small but non-zero.

“When devices were pooled, agreement degraded with darker pigmentation within the Fitzpatrick III–V range: the Bland–Altman dispersion shown in Figure 4 was broader at Fitzpatrick V and systematic bias was small but non-zero”

May combine two claims.

Device MAEs of approximately 2-4 bpm (Apple 2.51, Garmin 2.93, Samsung 3.38, Fitbit 3.89) fall within the ±5 bpm/±10% tolerance typically used as a pass criterion in the ANSI/CTA-2065 consumer wearable testing standard, and MAPE values of 2-4% are comparable to or narrower than beat-to-beat variability across repeated ECG readings at rest.

“the device MAEs of approximately 2–4 bpm in the present cohort (Apple 2.51, Garmin 2.93, Samsung 3.38, Fitbit 3.89 bpm; Table 1) fall within the ±5 bpm/±10% tolerance typically used as a pass criterion in the ANSI/CTA-2065 consumer wearable testing standard, and the corresponding MAPE values of 2–4% (Table 1) are comparable to or narrower than the beat-to-beat variability observed across repeated”

• AssumptionRests on an assumption: Beat-to-beat variability across repeated ECG readings at rest is a valid benchmark for MAPE.

May combine two claims.

A minority of participants disproportionately occupied the error tails; in the sample, several of these cases had higher BMI and darker skin tones.

“A minority of participants disproportionately occupied the error tails. In our sample, several of these cases had higher BMI and darker skin tones.”

May combine two claims.

In a single-site sample of Hispanic adults with Fitzpatrick skin types III-V, commercial wrist-PPG devices showed clinically acceptable agreement with ECG-derived heart rate during low-motion recumbent cycling, with a small-to-moderate systematic bias and greater error dispersion among participants with higher BMI and darker pigmentation.

“In a single-site sample of Hispanic adults with Fitzpatrick skin types III–V, commercial wrist-PPG devices showed clinically acceptable agreement with ECG-derived HR during low-motion recumbent cycling, with a small-to-moderate systematic bias and greater error dispersion among participants with higher BMI and darker pigmentation.”

• AssumptionRests on an assumption: The ECG chest-strap criterion provides valid heart rate measurement for agreement analysis.

May combine two claims.

Melanin absorbs light in the visible spectrum, especially at green wavelengths, reducing signal penetration into the blood-containing dermis and attenuating the AC component of the PPG signal.

“increased melanin reduces signal penetration into the blood-containing dermis, attenuating the AC component of the PPG signal.”

• cites [10] Zonios, 2001, Skin melanin, hemoglobin, and light scattering properties can be quantitatively assessed in vivo using diffuse reflectance spectroscopy, J. Investig. Dermatol. sourceNot checked — no full text of the source
• cites [11] Fallow, 2013, Influence of skin type and wavelength on light wave reflectance, J. Clin. Monit. Comput. sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Wrist assignment for each device (left vs. right) was randomized across participants.

“Wrist assignment for each device (left vs. right) was randomized across participants.”

May combine two claims.

The models indicated substantial unaccounted for inter-individual variability, underscoring the importance of accounting for participant-level nesting in heart-rate error analyses.

“suggesting substantial unaccounted for inter-individual variability and underscoring the importance of accounting for participant-level nesting in HR error analyses.”

May combine two claims.

When skin tone and obesity co-occur, the combined effect is more severe; Ajmal et al. noted a compounded degradation in signal quality when simulating a type VI skin tone with BMI > 40, with AC/DC ratio dropping more than 60%.

“Ajmal et al. [12] noted a compounded degradation in signal quality when simulating a type VI skin tone with BMI > 40, particularly in devices with shallow optical penetration depth like the Fitbit Versa 2. The AC/DC ratio dropped more than 60%”

• cites [12] Rodriguez, 2021, Monte Carlo analysis of optical heart rate sensors in commercial wearables: The effect of skin tone and obesity on the photoplethysmography (PPG) signal, Biomed. Opt. Express sourceNot checked — no full text of the source

Classified as probably not a substantial claim. May combine two claims.

Nonparametric statistical methods were prioritized to ensure robustness and interpretability and to accommodate the unequal cell sizes across Fitzpatrick groups.

“Accordingly, nonparametric statistical methods were prioritized to ensure robustness and interpretability and to accommodate the unequal cell sizes across Fitzpatrick groups”

Classified as probably not a substantial claim.

For the Fitzpatrick subgroup analyses, device-participant combinations were treated as independent observations in nonparametric tests.

“For the Fitzpatrick subgroup analyses, device–participant combinations were treated as independent observations in nonparametric tests.”

Classified as probably not a substantial claim.

The results were verified using a linear mixed-effects model with participant as a random effect.

“because participants contributed up to four devices, we verified the results using a linear mixed-effects model with participant as a random effect.”

May combine two claims.

The heart-rate slopes were small (<0.01 units per bpm) and varied by device, so they are interpreted as modest, device-specific patterns rather than a generalizable improvement with intensity.

“These slopes were small (<0.01 units per bpm) and varied by device, so we interpret them as modest, device-specific patterns rather than a generalizable improvement with intensity.”

• AssumptionRests on an assumption: The K5 metabolic system heart rate accurately indexes exercise intensity.

May combine two claims.

The heart-rate slopes were statistically significant but very small in magnitude (e.g., b = −0.0094 for √APE, explaining ~8% of variance), so improvements in accuracy with higher heart rate should be considered modest.

“Although the slopes were statistically significant, their magnitudes were very small (e.g., b = −0.0094 for √APE, explaining ~8% of variance), so improvements in accuracy with higher heart rate should be considered modest.”

• AssumptionRests on an assumption: The K5 metabolic system heart rate accurately indexes exercise intensity.

Probably not a substantial claim: may not pass S3 (load-bearing).

Participants were randomly assigned to wrist-placement configurations for the wearable devices.

“Participants were randomly assigned to wrist-placement configurations for the wearable devices”

Probably not a substantial claim: may not pass S3 (load-bearing).

All data are provided at OSF.

“All data are provided at OSF: https://osf.io/9pzr2/overview?view_only=575bd3c163464781a0b600aee6a4d6c6.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Sex differences in subcutaneous fat distribution may influence the light path length and scattering behavior.

“Sex differences in subcutaneous fat distribution may also influence the light path length and scattering behavior [20].”

• cites [20] Freedman, 2013, A comparison of the Slaughter skinfold-thickness equations and BMI in predicting body fatness and cardiovascular disease risk factor levels in children, Am. J. Clin. Nutr. sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

The recruitment target was increased to 58 individuals to mitigate potential data loss due to equipment malfunction or invalid sensor recordings.

“To mitigate potential data loss due to equipment malfunction or invalid sensor recordings, the recruitment target was increased to 58 individuals.”

Probably not a substantial claim: may not pass S3 (load-bearing).

All procedures were preregistered on OSF.

“All procedures were preregistered on OSF: https://osf.io/9pzr2/overview?view_only=575bd3c163464781a0b600aee6a4d6c6.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Preliminary checks confirmed that wrist side did not influence outcomes (ps > 0.05).

“Preliminary checks confirmed that wrist side did not influence outcomes (ps > 0.05).”

May combine two claims.

Heart-rate-dependent reductions in error were device-specific: Samsung showed consistent, though modest, improvements across all error metrics; Apple showed more isolated effects; Garmin showed mixed effects; and Fitbit demonstrated the strongest heart-rate-related improvement in relative accuracy.

“Collectively, these findings suggest that HR-dependent reductions in error were device-specific, with Samsung showing consistent, though modest, HR-related improvements across all error metrics, while Apple showed more isolated effects, Garmin showed mixed effects, and Fitbit demonstrated the strongest HR-related improvement in relative accuracy.”

• AssumptionRests on an assumption: The K5 metabolic system heart rate accurately indexes exercise intensity.

Classified as probably not a substantial claim. May combine two claims.

Interaction effects (BMI × Fitzpatrick; intensity × BF% × Fitzpatrick) are not confirmatory evidence of population-level moderation but are preregistered, exploratory patterns whose direction aligns with optical-physiological mechanisms hypothesized a priori.

“we emphasize interaction effects (BMI × Fitzpatrick in Section 3.2; intensity × BF% × Fitzpatrick in Section 3.7) not as confirmatory evidence of population-level moderation but as preregistered, exploratory patterns whose direction of effect aligns with the optical–physiological mechanisms hypothesized a priori.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

In free-living use, wrist rotation, variable contact pressure, and ambient light will introduce additional noise and could amplify both overall error and phenotype-linked disparities.

“In free-living use, wrist rotation, variable contact pressure, and ambient light will introduce additional noise and could amplify both overall error and phenotype-linked disparities.”

Classified as probably not a substantial claim.

All wearable streams were synchronized to an ECG chest-strap criterion.

“synchronized all wearable streams to an ECG chest-strap criterion”

• AssumptionRests on an assumption: The ECG chest-strap criterion provides valid heart rate measurement for agreement analysis.

May combine two claims.

Garmin's heart-rate error was highly sensitive to the combined influence of adiposity, pigmentation, and exercise intensity; individuals with higher BF% and darker skin tones exhibited greater directional overestimation at higher intensities, but simultaneously experienced improvements in absolute and relative error.

“These results suggest that Garmin’s HR error was highly sensitive to the combined influence of adiposity, pigmentation, and exercise intensity. Specifically, individuals with higher BF% and darker skin tones exhibited greater directional overestimation at higher intensities, but simultaneously experienced improvements in absolute and relative error”

May combine two claims.

This was a single-site study of self-identified Hispanic adults with Fitzpatrick skin types III-V; Fitzpatrick types I, II, and VI were not represented, and the Fitzpatrick V subgroup (n = 4) was small.

“this was a single-site study of self-identified Hispanic adults with Fitzpatrick skin types III–V; Fitzpatrick types I, II, and VI were not represented, and the Fitzpatrick V subgroup (n = 4) was small.”

Classified as probably not a substantial claim. May combine two claims.

Findings should not be generalized to Fitzpatrick I-II or VI, to non-Hispanic populations, to free-living contexts, or to pediatric or older adult populations without external replication.

“Findings should therefore not be generalized to Fitzpatrick I–II or VI, to non-Hispanic populations, to free-living contexts, or to pediatric or older adult populations without external replication.”

May combine two claims.

The exclusion of Fitzpatrick I-II and VI was a pragmatic scope decision driven by power (an a priori MANOVA calculation targeting 52-58 participants precluded adequate cell sizes across all six Fitzpatrick types), a fixed budget, and a local volunteer pool concentrated in mid-range Fitzpatrick tones.

“The exclusion of Fitzpatrick I–II and VI was a pragmatic scope decision driven by power (an a priori MANOVA calculation targeting 52–58 participants precluded adequate cell sizes across all six Fitzpatrick types), a fixed budget that capped participant incentives, and a local volunteer pool concentrated in mid-range Fitzpatrick tones”

May combine two claims.

Device-specific analyses yielded fewer skin-tone effects, and the cell size of the Fitzpatrick V subgroup was small, warranting cautious interpretation.

“Device-specific analyses yielded fewer skin-tone effects, and the cell size of the Fitzpatrick V subgroup was small, warranting cautious interpretation.”

May combine two claims.

The clinical meaningfulness of these point estimates is dependent on intended use: acceptable for zone-based training and routine monitoring, potentially consequential for rigid clinical thresholds or long-duration energy balance calculations, and especially relevant where systematic bias clusters within phenotypic subgroups.

“The clinical meaningfulness of these point estimates is therefore best read as dependent on intended use: acceptable for zone-based training and routine monitoring, potentially consequential for rigid clinical thresholds or long-duration energy balance calculations, and especially relevant where systematic bias clusters within phenotypic subgroups as observed here.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The protocol minimized gross motion with recumbent cycling at a standardized cadence and arm stabilization.

“We minimized gross motion with recumbent cycling at a standardized cadence and arm stabilization”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

A recumbent, low-motion context approximates a best-case optical environment; however, it may not benefit from accelerometry-informed algorithms.

“A recumbent, low-motion context approximates a best-case optical environment; however, it may not benefit from accelerometry-informed algorithms.”

May combine two claims.

Because the Fitzpatrick V stratum is underpowered, any interaction coefficient that depends on Fitzpatrick V cells is tentative and will only become confirmatory when replicated in larger, purposively stratified samples; the Bland-Altman visualization is leaned on as the more defensible evidentiary base within this cohort.

“Because the Fitzpatrick V stratum is underpowered, any interaction coefficient that depends on Fitzpatrick V cells is necessarily tentative and will only become confirmatory when replicated in larger, purposively stratified samples; we flag these interactions accordingly and lean on the Bland–Altman visualization in Figure 4 and Figure 6 as the more defensible evidentiary base within this cohort.”

May combine two claims.

Fitbit devices exhibited biased heart-rate estimation as a function of skin tone and intensity, but also demonstrated improvements in absolute and relative accuracy under higher-intensity conditions for individuals with higher pigmentation.

“Together, these results suggest that while Fitbit devices exhibited biased HR estimation as a function of skin tone and intensity, they also demonstrated improvements in absolute and relative accuracy under higher-intensity conditions for individuals with higher pigmentation”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Nonparametric tests were pre-specified and suited to bounded, skewed error distributions.

“pre-specified nonparametric tests suited to bounded, skewed error distributions”

May combine two claims.

Commercial wearables work well on average under favorable conditions, while still exhibiting phenotypic sensitivity that can matter in specific contexts.

“These observations reinforce that commercial wearables work well on average under favorable conditions, while still exhibiting phenotypic sensitivity that can matter in specific contexts.”

May combine two claims.

Current commercial systems work well on average but remain vulnerable to errors in certain individuals, which are not random with respect to phenotypic features.

“we highlight that current commercial systems work well on average but remain vulnerable to errors in certain individuals, which are not random with respect to phenotypic features.”

May combine two claims.

Apple's heart-rate error was moderated by both skin tone and body composition, with combined effects of higher exercise intensity, BF%, and pigmentation tending to increase directional bias but reduce relative error.

“Taken together, these results indicate that Apple’s HR error was moderated by both skin tone and body composition, with combined effects of higher exercise intensity, BF%, and pigmentation tending to increase directional bias, but reduce relative error”

May combine two claims.

Intensity-linked improvements are modest rather than decisive, but suggest that PPG systems may perform comparatively better when stroke volume and pulsatility increase.

“We therefore interpret intensity-linked improvements as modest rather than decisive, but they do suggest that PPG systems may perform comparatively better when stroke volume and pulsatility increase [2,29].”

• cites [2] Garet, 2005, Estimating relative physical workload using heart rate monitoring: A validation by whole-body indirect calorimetry, Eur. J. Appl. Physiol. sourceNot checked — no full text of the source
• cites [29] Maeda, 2011, Relationship between measurement site and motion artifacts in wearable reflected photoplethysmography, J. Med. Syst. sourceNot checked — no full text of the source

May combine two claims.

Given modest subgroup sizes and collinearity between Fitzpatrick 5 and high BMI in these individuals, the observation that error-tail cases had higher BMI and darker skin tones is hypothesis-generating rather than confirmatory.

“However, given the modest subgroup sizes and collinearity between Fitzpatrick 5 and high BMI in these individuals, this observation is hypothesis-generating rather than confirmatory.”

Classified as probably not a substantial claim. May combine two claims.

Mixed-effects models respected within-person nesting, allowing separation of systematic influences from participant-level heterogeneity without overfitting device noise.

“We then complemented those tests with mixed-effects models that respected within-person nesting, allowing us to separate systematic influences from participant-level heterogeneity without overfitting device noise.”

May combine two claims.

Samsung devices tended to underestimate heart rate during high-intensity exercise in individuals with both higher adiposity and darker skin; however, both absolute and relative error increased in these same individuals, indicating that precision and percentage accuracy deteriorated.

“These findings suggest that Samsung devices tended to underestimate HR during high-intensity exercise in individuals with both higher adiposity and darker skin. However, both absolute and relative error increased in these same individuals, indicating that although the device’s bias trended downward, its precision and percentage accuracy deteriorated”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The trial was short and intensity-segmented, so longer bouts may reveal thermal or vasomotor drift that the protocol could not capture.

“The trial was short and intensity-segmented, so longer bouts may reveal thermal or vasomotor drift that our protocol could not capture.”

May combine two claims.

Skin-tone representation was limited (Fitzpatrick V underpowered; type VI absent), which may limit the detectability of pigmentation effects.

“Skin-tone representation was limited (Fitzpatrick V was underpowered; type VI was absent), which may limit the detectability of pigmentation effects.”

Classified as probably not a substantial claim. May combine two claims. The quote only partly states this claim.

The two- and three-way interaction terms involving Fitzpatrick skin tone are preregistered exploratory analyses and are flagged as hypothesis-generating rather than confirmatory due to small participant-level cell sizes (Fitzpatrick V n = 4).

“we emphasize that these interactions are preregistered exploratory analyses”

• Also stated in: § 3.7. Hierarchical Linear Modeling · plain

  “We therefore flag significant two- and three-way interaction coefficients involving Fitzpatrick skin tone as hypothesis-generating rather than confirmatory”

May combine two claims.

These magnitudes are meaningfully smaller than the width of typical aerobic training zones and therefore unlikely to flip a user between zones during recreational training.

“These magnitudes are meaningfully smaller than the width of typical aerobic training zones (e.g., moderate 64–76% vs. vigorous 77–95% HRmax) and are therefore unlikely to flip a user between zones during recreational training.”

Probably not a substantial claim: may not pass S2 (checkable).

For practitioners and end-users, these devices are suitable for most training decisions, especially when corroborated by perceived exertion and performance cues.

“Translationally, for practitioners and end-users, these devices are suitable for most training decisions, especially when corroborated by perceived exertion and performance cues.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Although mechanisms driving the phenotypic effects (optical attenuation by melanin; altered dermal properties with higher adiposity) operate independently of self-identified ethnicity, similar directional patterns would be expected in non-Hispanic cohorts only where Fitzpatrick and BMI distributions overlap the study's, with magnitudes that may differ.

“Although the mechanisms driving the phenotypic effects observed here (optical attenuation by melanin; altered dermal properties with higher adiposity) operate independently of self-identified ethnicity, similar directional patterns would be expected in non-Hispanic cohorts only where Fitzpatrick and BMI distributions overlap ours, with magnitudes that may differ”

Probably not a substantial claim: may not pass S2 (checkable).

Caution is warranted when applying rigid thresholds to individuals with higher adiposity and darker pigmentation.

“However, caution is warranted when applying rigid thresholds (e.g., HR-defined clinical alerts) to individuals with higher adiposity and darker pigmentation.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The study did not manipulate firmware or algorithms, and versions may have evolved during data collection, constraining the specificity of device-level inferences.

“Lastly, we did not manipulate firmware or algorithms, and versions may have evolved during data collection, constraining the specificity of device-level inferences.”

May combine two claims.

These findings support the use of wrist-PPG heart rate for most everyday training decisions within this cohort while indicating phenotype-linked agreement drift that warrants targeted evaluation in broader samples.

“These findings support the use of wrist-PPG HR for most everyday training decisions within this cohort while indicating phenotype-linked agreement drift that warrants targeted evaluation in broader samples”