Validation of the Deep Health® Device for body composition blood pressure and heart rate

Abstract Background. Multimodal physiological devices require validation against references appropriate to each reported output. The published Deep Health® standardization framework extends beyond dual-energy X-ray absorptiometry (DXA) soft-tissue estimates by using body-density and whole-body potassium methods for fat and skeletal muscle. Objective. To quantify agreement between Deep Health® Device outputs and body-composition reference methods, and between device blood pressure and heart rate and simultaneously acquired manual measurements, in two separate cohorts. Methods. A secondary analysis included 4,144 participants with paired body-fat percentage, skeletal-muscle mass and total-body bone mineral density measurements, and a separate cohort of 483 participants with paired systolic blood pressure (SBP), diastolic blood pressure (DBP) and heart rate (HR). Body-composition references were BOD POD air-displacement plethysmography, whole-body potassium-derived muscle mass and DXA bone density. Manual BP and HR measurements were taken while device measurements were also being acquired. Bias, Bland–Altman limits of agreement, mean absolute error (MAE), root mean square error and Lin concordance were calculated. Cohorts were analyzed separately without subject-ID linkage. Results. Body-fat bias was +0.038 percentage points, with 95% limits of −0.787 to +0.862; muscle-mass bias was −0.022 kg, with limits of −0.782 to +0.737; and bone-density bias was +0.00075 g/cm², with limits of −0.02294 to +0.02443. Concordance coefficients were 0.9986, 0.9987 and 0.9957, respectively. SBP, DBP and HR biases were −0.011 mmHg, −0.015 mmHg and +0.044 bpm; corresponding MAEs were 0.870 mmHg, 0.593 mmHg and 0.344 bpm. All BP differences were within 5 mmHg. HR errors varied with measurement magnitude. Conclusions. The supplied paired datasets show close agreement for six physiological outputs within the conditions and ranges studied. These findings provide quantitative support for an integrated, endpoint-specific validation program. They do not establish direct body-protein accuracy, longitudinal stability or equivalence across device generations. Independence from model development and a discrepancy in historical BMI eligibility require clarification before submission. Keywords: Deep Health® Device; body composition; bone mineral density; air-displacement plethysmography; total body potassium; skeletal muscle; blood pressure; heart rate; agreement