In October 2025, two pieces of evidence pointing in opposite directions were published almost simultaneously, together answering a question many people are paying real money to ask: Can buying a smartwatch that measures blood pressure and encourages physical activity actually lower your blood pressure? The first comes from a systematic review and meta-analysis published in *Hypertension Research*, synthesizing 21 randomized controlled trials—enrolling adults with hypertension, type 2 diabetes, or overweight/obesity—and testing interventions using pedometers, activity trackers, or smartwatches to promote physical activity over durations of 12 to 48 weeks. The conclusion is starkly clear: no statistically significant improvement was observed in systolic or diastolic blood pressure at any time point (12, 24, or 48 weeks), nor in weight, fasting glucose, or HbA1c. The second piece of evidence comes from the IMPACT-BP trial, published the same month in the *New England Journal of Medicine* (*NEJM*), with results pointing in the exact opposite direction—and with substantially larger effect sizes. Reading the two side by side does not yield the answer to “Do wearables work?” but rather reveals a far more critical distinction.
## Devices Acting Alone: 21 Trials, Zero Effect
First, let’s detail the negative meta-analysis—because it defines the lower bound of what “devices acting alone” can achieve. All 21 trials shared a common design: provide a device, promote activity, and route data directly to the patient. The authors searched major medical databases and included trials with follow-up durations ranging from 12 to 48 weeks; populations and device types varied—but conclusions on blood pressure endpoints were highly consistent: no significant improvements in systolic or diastolic blood pressure at 12, 24, or 48 weeks. The authors’ wording is direct: “Direct intervention based solely on monitoring lacks sufficient evidence to establish clinically meaningful blood pressure improvement.” They conclude that future efforts must integrate wearable technology into comprehensive care models featuring “targeted feedback + education + behavioral support”—not leave monitoring to operate in isolation.
It must be clarified that this negative conclusion applies specifically to physiological endpoints such as blood pressure and weight—and cannot be casually extended to claim “wearables are ineffective for behavior change,” since behavioral endpoints were not systematically assessed in this meta-analysis. Yet for the most common purchasing motivation—“Will buying a watch lower my blood pressure?”—the evidence is unequivocal: No.
## People Behind the Data: Systolic BP Reduced by 7.9–9.1 mm Hg
Now consider the positive evidence. IMPACT-BP was a three-arm, open-label, randomized trial conducted in rural South Africa, enrolling 774 adults (mean age 62 years; 76.0% female; 13.6% with diabetes; 46.5% living with HIV) who were randomized to one of three care arms: standard clinic-based care; home-based care—patients self-measured blood pressure, community health workers visited homes to collect data and deliver medications, and remote nurses guided treatment decisions via a mobile app; and enhanced home-based care—identical to the prior arm, except that blood pressure readings were automatically transmitted, eliminating manual recording.
At 6 months, systolic blood pressure in the two home-based care arms was 7.9 mm Hg (95% CI −10.5 to −5.3) and 9.1 mm Hg (−11.7 to −6.4) lower than in the standard-care arm, both P < 0.001. Hypertension control rates rose from 32.5% in the standard-care arm to 57.4% and 61.3% in the two home-based arms, corresponding to relative risks of 1.76 and 1.89. These improvements persisted at 12 months. Rates of serious adverse events (2.7%) and death (1.0%) were similar across all three arms, and retention exceeded 95% in both intervention arms. This is not merely “prettier numbers”—it represents nearly doubled care outcomes.
## Automation Adds Only 1.2; Closed Loops Add 7.9
The most analytically valuable detail in IMPACT-BP lies in the difference between the two home-based care arms. Their sole distinction was automated transmission of blood pressure readings: the point estimate for systolic reduction differed by just 1.2 mm Hg (−9.1 vs. −7.9), and control-rate relative risks were 1.89 vs. 1.76. Note that this is an arithmetic comparison of point estimates—not a formal statistical test between the two intervention arms; their respective 95% confidence intervals overlapped substantially when compared against standard care—so the added benefit of automation remains suggestive, not definitive.
Yet even by point estimate alone, this 1.2 mm Hg gain pales beside the leap—from no closed-loop care to closed-loop care—that yielded 7.9–9.1 mm Hg reductions and near-doubling of control rates. Let the trial’s own structure speak: sensor and connectivity upgrades are second-order variables; the first-order variable is whether a person capable of adjusting medications, delivering drugs, and conducting follow-ups stands behind the data.
The U.S. contributes a structurally analogous piece of the puzzle. The TOUCHED trial (published in *JAMA Cardiology*, July 2025) enrolled 574 patients discharged from an urban academic medical center emergency department with elevated blood pressure (≥140/90 and ≤180/110 mm Hg; mean age 51.1 years; 56% female; 72% Black; 20% Hispanic/Latino). Participants were randomized to receive either usual discharge instructions or a bundled intervention: brief hypertension counseling by clinical pharmacists or advanced practice nurses; daily self-monitoring using a smartphone-linked blood pressure monitor; behavior-change text messages; and referral to primary care. Among the 413 participants with 6-month blood pressure data, the intervention group achieved a significantly greater systolic reduction than the usual-care group—by 4.9 mm Hg (95% CI 0.8–9.0; P = .02); however, the proportion achieving blood pressure control (42.9% vs. 36.9%) did not differ significantly. Its component breakdown is revealing: devices accounted for only one-quarter of the intervention—the rest consisted of people (pharmacist/nurse counseling), channels (text messaging), and linkages (primary care referral). Its limitations are equally transparent: only 413 of 574 enrollees contributed 6-month data (~28% missing), it was single-center, its multi-component package precludes isolating individual contributions, and the harder endpoint—“control”—was not statistically significant.
## The Most Counterintuitive Finding: Greatest Effects Where Resources Are Scarcest
When these three bodies of evidence are viewed together, another cross-trial pattern emerges: the largest effects occurred precisely where resources were most limited—rural South Africa (IMPACT-BP) and the U.S. safety-net emergency department population (TOUCHED). This is cross-trial interpretation—not head-to-head testing—but the direction is consistent: these populations lack not devices, but access to care. When home-based care packages nurse-led decision-making, medication delivery, and follow-up visits directly to patients’ doorsteps, outcomes surpass those seen when devices are deployed across heterogeneous populations. Technology is not the variable; organizational design is.
## Three Takeaways You Can Use
First, for consumers: view your smartwatch as a “sensor without a closed loop.” Twenty-one randomized trials indicate it is unlikely to lower your blood pressure on its own. What truly adds value is connecting its data into a channel where someone reviews it, adjusts medications, and conducts follow-ups—your family physician or a remote care program. Before buying, ask: Who receives the readings—and who has authority to act on them?
Second, for health systems and manufacturers: IMPACT-BP’s cost structure suggests investment priorities should focus first on “the people behind the data”—nursing decision capacity, medication delivery, and follow-up—rather than on devices and automation. When procuring wearables, the question “Who receives the readings—and who has authority to act on them?” predicts effectiveness more reliably than device accuracy.
Third, for readers of evidence: citing any single study permits telling whichever story you prefer—manufacturers might cite only TOUCHED’s “smart BP monitor lowered systolic BP,” while skeptics might cite only the meta-analysis’s “all 21 trials negative.” Only when all three are viewed side-by-side do mechanisms become visible: data collection ≠ intervention; a closed loop = data → human review → treatment adjustment → medication access → follow-up; automation mitigates loop inefficiencies—not the loop itself.
None of the three trials followed participants beyond 12 months, so long-term adherence decay remains unknown; IMPACT-BP used an open-label design, and its combined nurse–community health worker model demands high personnel investment—costs in other health systems remain unmeasured. But the direction is already unmistakable. A critical question now arises: Consumer-grade wearables have sold hundreds of millions of units—but the authority to “act upon your data” is rarely explicitly assigned to any care provider. Who should fill that gap?
Source: Home-Based Care for Hypertension in Rural South Africa (IMPACT-BP). *N Engl J Med* 2025. https://doi.org/10.1056/NEJMoa2509958