Not the Ring, the Outcome Loop: What Metabolic Sensing Cannot Prove for MBG

MBG Watch · 2026-08-30

The premise

The new signal is real, but narrow. In July 2026, UC San Diego reported a smart-ring prototype that can monitor several sweat biomarkers, including glucose, ketones, vitamin C, uric acid, lactate, and alcohol, and send readings to a phone app. The linked Nature Communications article describes a fully integrated ring that reads biomarkers from passively extracted sweat, supports about 12 hours of continuous monitoring, and was validated against commercial blood meters and glucose monitors in healthy and type 1 diabetic participants.

That is useful science. It is not evidence that a national feeding program has improved child nutrition.

MBG is operating at a different scale and with a different duty. Badan Gizi Nasional has described 2026 targets up to 82.9 million beneficiaries, including students, pregnant women, breastfeeding mothers, toddlers, and remote or underserved areas. It has also reported early-2026 operational figures of more than 21,000 SPPG units and about 59.86 million reached beneficiaries. At that scale, the question is not whether a device can produce a plausible metabolic trace. The question is whether any trace can be interpreted lawfully, clinically, fairly, and modestly enough to support a public claim.

MBG Watch has already argued this boundary in adjacent settings. "Performance Is Not Proof" separated attendance, focus, and enthusiasm from durable child benefit. "Earlier Care Is Not Earlier Proof" argued that delivery, safety, and health outcomes for mothers, toddlers, and 3T areas must not be collapsed into one dashboard. "Seen Without Being Watched" set a privacy boundary around children and health data. "Not the Sensor, the Measurement Chain" said that no small-device reading becomes evidence until the chain around it is inspectable. The same standard applies here.

A metabolic sensor may describe a moment in a body. MBG needs evidence of nutrition outcomes across a population.

What the evidence supports

The evidence supports three careful statements.

First, biochemical wearables are becoming more capable. The UC San Diego ring is not a mere step counter with a health label attached. It attempts continuous biochemical sensing from sweat and compares some readings with accepted glucose-monitoring references. This matters because future vendors may soon offer MBG-like programs cheaper, softer, and more continuous biomarker monitoring than laboratory testing.

Second, sweat sensing still has interpretation limits. A review of wearable sweat sensors notes that some blood analytes, including glucose, lactate, and ethanol, have been reported to correlate with sweat levels, but also says sweat measurement still has data-reliability issues and that secretion mechanisms for some analytes remain unknown. In practice, a reading can be shaped by sweat rate, skin contact, hydration, temperature, device calibration, placement, battery state, and the algorithm that turns a weak biochemical signal into a human-facing number.

Third, nutrition outcomes already have a lower-risk measurement vocabulary. WHO describes child-growth assessment through anthropometric indicators such as length/height-for-age, weight-for-age, weight-for-length or weight-for-height, and body mass index-for-age. A school-feeding systematic review and meta-analysis in low- and middle-income countries included education outcomes, anthropometry, micronutrient status including hemoglobin and anemia, cognition, and morbidity; it found modest average gains in height, weight, and attendance over 12 months, while still calling for stronger research. These are not perfect measures. But they are closer to the public-health question MBG must answer than a private stream of raw metabolic traces.

What a biomarker is not

A biomarker is not a denominator. It does not tell the public how many eligible children were reached, how often meals were actually consumed, how many meals met specification, how many children were missed, or whether the sample represents the population.

A biomarker is not a baseline. A glucose, ketone, lactate, vitamin, hemoglobin, or growth reading after meal delivery has little public meaning unless the program knows where the person or cohort started, what would have happened without the intervention, and what else changed at the same time.

A biomarker is not consent. The fact that a device is small, soft, or marketable does not make it ethically light. In Indonesia, health data and records, biometric data, genetic data, and children’s data fall within the category of specific personal data under the Personal Data Protection Law as summarized by DLA Piper. Pregnant status, breastfeeding status, child identity, metabolic traces, and location-linked attendance patterns are not ordinary administrative fields.

A biomarker is not clinical care. An abnormal reading may require interpretation, counseling, referral, retesting, or protection from stigma. If MBG collects a signal it cannot responsibly act on, it has created anxiety and liability without a care pathway.

A biomarker is not proof of program impact. It may be useful in a clinical study. It may help a clinician understand an individual risk. It may help a researcher test a narrow mechanism. But a national accountability claim needs a method, a denominator, a comparator, uncertainty bounds, and a correction trail.

The minimum outcome loop before any device data counts

If MBG ever studies metabolic sensing, the burden should be high and explicit. The loop should include at least nine gates.

1. Baseline: the cohort’s starting nutrition and health status must be recorded before the intervention being evaluated.

2. Denominator: the public must know who was eligible, who was invited, who consented, who refused, who dropped out, who had missing readings, and who was excluded from analysis.

3. Consent and legal basis: consent must be informed, voluntary, revocable, and separated from meal entitlement. A child or family should never have to surrender intimate health data to receive food.

4. Comparator: the evaluation must identify a credible comparison group or design. Otherwise the device only records change, not cause.

5. Clinical interpretation: every biomarker used must have a documented clinical meaning for the population being measured, including children, pregnant women, breastfeeding mothers, toddlers, and relevant local contexts.

6. Follow-up and referral: abnormal or concerning readings must lead to a defined care path, not a dashboard flag.

7. Uncertainty: device error, calibration drift, sweat-rate effects, missingness, skin and environment bias, and algorithmic interpretation must be reported with the result.

8. Correction: participants must have a way to challenge, correct, or delete data where law and ethics require it.

9. Retention and deletion: raw traces should not be kept because they might be useful someday. Retention should be short, justified, and auditable.

Without these gates, a metabolic signal should not support any public claim that MBG improved nutrition.

What should not be collected for routine accountability

Some data should likely never enter routine MBG accountability systems.

Raw metabolic traces from children should not be collected to prove program performance. Individual pregnancy or breastfeeding health status should not be exposed outside clinical care. Precise location histories, attendance-linked movement patterns, biometric identifiers, device IDs, and child-level app telemetry should not become the price of public nutrition measurement. Missing or abnormal readings should never be treated as fraud, family failure, or grounds for exclusion.

This is not hostility to measurement. It is respect for proportionality. The ASIK support materials describe Posyandu recording as a front line for detecting stunting and note why better, more systematic recording matters: manual books, repeated data entry, and input errors can distort nutrition status. That is a better starting point for MBG accountability than a new stream of intimate device data. Improve the records that already connect to care before creating a parallel surveillance layer.

Lower-risk measurements that should come first

The lower-risk path is not data-poor. It is better ordered.

MBG should first publish district- or cohort-level anthropometry where appropriate; anemia or micronutrient surveillance where clinically justified; ASIK- or Posyandu-linked aggregate indicators; meal exposure by beneficiary group; safety incidents; completed referrals; kitchen operating status; and credible comparison groups. For school-aged children, attendance and classroom indicators may be useful context, but they should not be renamed as nutrition outcomes. For mothers, toddlers, and 3T beneficiaries, delivery success must be separated from health effect and from safety.

The public record should answer simple questions before intimate ones. How many eligible beneficiaries received meals? How often? What was served? Were meals safe? Which cohorts improved on accepted nutrition indicators? What was the uncertainty? Which areas did not improve? What changed after the finding?

That is enough to learn. It is not enough to watch.

The least-harm position

MBG should not use wearable metabolic sensing as a shortcut to claim impact.

If sweat, glucose, ketone, vitamin, lactate, anemia, or other biomarker tools are ever studied, they belong first in tightly governed clinical or research pilots with independent ethics review, clear consent, careful sampling, clinical referral, and public reporting only in aggregate. They should not be routine validation tools. They should not decide eligibility, payment, school compliance, SPPG performance, household blame, or public proof.

The better standard is the same one MBG Watch has applied to other tempting technologies: not the sensor, the measurement chain. For nutrition outcomes, that chain is baseline, denominator, comparator, clinical interpretation, privacy boundary, uncertainty, and correction.

A ring may one day help a person understand a body. MBG must prove that meals improve children’s and mothers’ wellbeing without turning their bodies into the dashboard.

What I am uncertain about

I am uncertain how close the UC San Diego ring or similar devices are to large-scale clinical deployment; the reported work is promising, but public program readiness requires more than prototype validation.

I am uncertain which biomarker, if any, MBG authorities or vendors are currently considering. This analysis treats wearable metabolic sensing as a plausible future pressure, not as a confirmed procurement plan.

I am uncertain how ASIK, Posyandu, school, and MBG records will be linked in practice. That linkage could improve aggregate outcome learning if minimized and governed well; it could also increase risk if identifiers and health fields are joined without strict purpose limits.

The practical conclusion does not depend on those uncertainties. Until MBG can show a privacy-preserving outcome loop, metabolic sensing should remain research infrastructure, not public proof.