Biomarker Signatures of Underfueling: Sleep, HRV, RHR, and Their Limits

Your ring tells you your HRV is down 18 percent over the last month. Your resting heart rate is up 6 bpm. Your sleep score has been below your baseline four nights out of seven.

Is your body telling you you’re underfueling? Or is it telling you something else entirely?

This post walks signal by signal through the biomarkers wearables can show and what the published literature says about their relationship to low energy availability. For each signal, we cover the directionality of change under sustained low EA, the confounders that produce similar changes, and what the honest interpretation is.

This is a spoke in the energy availability and RED-S pillar. For the overall framing, start there. This post is the mechanical breakdown.

What “Signature” Means

A physiological signature is a consistent pattern of change across one or more biomarkers that reflects an underlying state. For low energy availability, the signature in the research literature comprises a cluster of changes: autonomic shift toward lower HRV and altered RHR, suppressed reproductive function, decreased metabolic rate, disrupted sleep, stalled or regressing training response, and changes in mood and cognition.

Not every low-EA athlete shows every element. Not every element, when present, is caused by low EA. What the framework proposes is that when the signatures cluster, and when they persist while known confounders are not active, the probability that low EA is contributing rises.

The discipline of signature recognition is in the ensemble, not any single signal.

Heart-Rate Variability (HRV)

Directionality under low EA. Sustained low energy availability is associated with reduced time-domain HRV (RMSSD, SDNN) in most studies. The proposed mechanism is a shift in autonomic balance — reduced parasympathetic tone combined with altered sympathetic drive. The shift can be gradual over weeks and more pronounced in athletes with longer durations of sub-threshold EA.

Published support. Stellingwerff and colleagues (2021) synthesized evidence for autonomic changes during RED-S-relevant conditions. Multiple smaller studies on endurance athletes during periods of low EA have shown modest but consistent HRV reductions. The effect size is typically 10–25 percent reductions in RMSSD from individual baseline, developing over several weeks.

Confounders that produce the same directional change:

• Luteal phase of the menstrual cycle lowers HRV in most women, reflecting progesterone-driven shifts in autonomic balance. The magnitude (often 10–20 percent from follicular baseline) overlaps exactly with the low-EA signal.
• Acute illness, particularly viral, sharply lowers HRV in the days before overt symptoms and during acute illness. Sometimes the HRV drop is the earliest sign.
• Alcohol consumption suppresses HRV for 24–48 hours after drinking, with dose-dependence.
• Acute training load spikes transiently lower HRV for 24–72 hours; in fitness terms this is the “fitness-fatigue” model playing out in the autonomic signal.
• Sleep deprivation lowers HRV roughly proportional to sleep debt accumulated.
• Air travel across time zones lowers HRV for several days.
• Acute psychological stress lowers HRV.

Honest interpretation. A sustained downward HRV trend relative to your own phase-matched baseline, over 2–3 weeks, with no obvious confounder active, and co-occurring with other signals in the underfueling cluster, adds evidence for the underfueling hypothesis. A 3-day HRV drop without context is essentially meaningless for EA inference. The more specific the HRV drop signal becomes (i.e., the more confounders can be ruled out), the more weight it should carry.

For the signal fundamentals, see what HRV is and how wearables measure it. For the composite-score context, see composite scores with confidence.

Resting Heart Rate (RHR)

Directionality under low EA. The literature on RHR during low EA is more mixed than on HRV. Early-stage or mild low EA often shows modest RHR elevation; advanced or severe low EA sometimes shows bradycardia, consistent with the metabolic suppression seen in prolonged cases. The direction of change depends on stage and severity.

Published support. The mixed picture is itself in the literature. Clinical RED-S case series document both elevated and reduced RHR in affected athletes, and the direction often correlates with duration of the condition rather than severity alone. In consumer surveillance, the elevated-RHR signal is the more common early signal.

Confounders:

• Luteal phase elevates RHR by 2–5 bpm in most women even with adequate fueling.
• Illness elevates RHR, often several days before overt symptoms and for a week or more during recovery.
• Alcohol elevates RHR for 24–48 hours post-consumption, dose-dependent.
• Acute training load elevates RHR for 1–3 days after hard sessions.
• Sleep debt elevates RHR.
• Dehydration elevates RHR.
• Environmental heat elevates RHR.
• Caffeine elevates RHR acutely.
• Anxiety elevates RHR.

Honest interpretation. A sustained RHR elevation above phase-matched baseline, persisting 2+ weeks, with no obvious confounder, is worth noticing. RHR is a simple metric — easy to measure, easy to understand, easy to track — which makes it tempting to overweight. Don’t. It’s non-specific enough that single-signal interpretation is unreliable.

Sleep Duration and Architecture

Directionality under low EA. Low EA has been associated with total sleep time reductions, increased sleep fragmentation, and reduced deep (slow-wave) sleep proportion. Athletes in documented low-EA periods frequently report subjectively worse sleep.

Published support. The sleep literature in RED-S is less robust than the autonomic literature. Studies are smaller, sleep measurement in free-living athletes is less standardized, and the association is less consistent across studies.

Confounders:

• Alcohol disrupts sleep architecture (suppresses REM, increases fragmentation) for 24–48 hours.
• Caffeine late in the day disrupts sleep onset and architecture.
• Late meals (particularly large ones) disrupt sleep.
• Environmental factors — room temperature, light exposure, noise.
• Travel and jet lag.
• Stress and rumination.
• Medication changes.
• Hormonal changes across the menstrual cycle affect sleep stages.
• Illness disrupts sleep.
• Late evening exercise can delay sleep onset.

Honest interpretation. Sleep is the signal most vulnerable to confounding. Dozens of normal-life factors disrupt sleep, most of which have nothing to do with energy availability. Sleep data is valuable as a corroborating signal in a cluster, not as a primary flag for low EA.

For sleep-signal fundamentals, see how to interpret your sleep score across devices.

Body Mass Trajectory

Directionality under low EA. Unintentional body mass reduction during training increases is a relatively high-specificity signal for low EA. When intake is insufficient to support training demands, the body preferentially draws on glycogen and fat stores, and over weeks this shows up in scale weight.

Published support. Body mass trajectory is used as a central input in clinical RED-S assessment. Unintentional weight loss is one of the flagged items on the IOC RED-S CAT 2 risk assessment tool.

Confounders:

• Intentional weight cuts — competition weight classes, intentional dieting — are not confounders of the underlying mechanism but are confounders of the “unintentional” framing. The body doesn’t know whether the cut is intentional; the physiological consequence of low EA is the same. Deliberate weight loss during training often produces low-EA states.
• Hydration shifts produce 1–2 kg swings on any given day. Individual scale readings are noisy.
• Glycogen depletion during a heavy training block reduces stored glycogen (and its bound water) by 1–2 kg without any actual fat mass change.
• GI contents vary by 0.5–1 kg day to day based on what and when you ate.
• Menstrual cycle produces fluid shifts that affect scale weight across the cycle.
• Illness often produces short-term weight loss from reduced intake.

Honest interpretation. Multi-week downward trends during training increases, especially when unintentional, are higher-specificity than most wearable signals. Single-week weight drops or single-day readings are dominated by confounders. For method detail, see body composition: DEXA vs smart scales vs calipers.

Menstrual Function

Directionality under low EA. The reproductive axis is highly sensitive to energy availability in women. The Loucks and Thuma research established the 30 kcal/kg FFM/day threshold for luteinizing hormone pulsatility disruption. Downstream effects include luteal-phase shortening, anovulatory cycles, oligomenorrhea, and eventually secondary amenorrhea.

Published support. This is the most-replicated endocrine signal in the low-EA literature and is a central component of the Female Athlete Triad framework.

Confounders:

• Hormonal contraceptives mask menstrual function as a signal — users have pharmacologically controlled cycles rather than naturally regulated ones, which makes this biomarker unusable for EA inference.
• Polycystic ovary syndrome (PCOS) produces irregular or absent cycles by different mechanisms.
• Thyroid disease can affect cycle regularity.
• Psychological stress can delay or suppress menstruation.
• Rapid body composition changes (not specifically low EA) can affect cycle regularity.
• Perimenopause alters cycle patterns.

Honest interpretation. Secondary amenorrhea (three or more consecutive missed periods in a previously regularly-menstruating athlete not on hormonal contraception and with no other explaining diagnosis) is one of the stronger signals. It warrants clinical evaluation, not a wearable flag. Oligomenorrhea (irregular periods) is less specific and still warrants clinical attention.

Strength and Training Adaptation

Directionality under low EA. Sustained low EA is associated with blunted strength gains, reduced endurance adaptation, and in some cases regressions. The hypothesized mechanism involves impaired protein synthesis and reduced glycogen availability compromising training stimulus.

Published support. Plausible and increasingly documented, but the noise in strength and performance data is high enough that attributing plateaus to EA specifically is difficult.

Confounders:

• Technique ceilings plateau many athletes.
• Programming errors — insufficient progressive overload, poor exercise selection.
• Sleep deprivation compromises adaptation.
• Overreaching temporarily suppresses performance.
• Illness and recovery.
• Aging produces gradual strength changes.
• Specific muscle recovery deficits — individual muscles have different adaptation rates; plateaus often reflect one muscle’s readiness rather than systemic.

Honest interpretation. A stalled strength progression is a weak signal on its own. In the context of other underfueling signatures, it adds to the pattern. Don’t read stalled PRs as a RED-S flag in isolation. For the training-load context, see adaptive training intelligence and what ACWR is.

Subjective Signals: Mood, Libido, Cold Intolerance

Directionality under low EA. RED-S clinical presentations include blunted mood, low libido (in both sexes), cold intolerance, reduced cognitive function, dry skin, constipation, and reduced resting body temperature. These are clinical findings from the RED-S literature.

Published support. Well-documented clinically. Hard to measure objectively with wearables — most of these require self-report or clinical examination.

Confounders:

• Depression produces most of these symptoms independently of EA.
• Thyroid dysfunction produces cold intolerance, constipation, fatigue, dry skin.
• Sleep deprivation produces mood and cognitive effects.
• Relationship or life stress affects mood and libido.
• Medication side effects.
• Perimenopause and menopause.
• Autumn and winter seasons in some populations (seasonal affective patterns).

Honest interpretation. Self-reported subjective signals are best used as corroborating evidence in a pattern-recognition context, via consistent logging over time. A single rough week on mood is expected variance. Sustained low-grade mood disturbance, persistent cold intolerance not explained by environment, and persistent low libido are worth noting and worth a clinic conversation.

The Compound Signal

The useful signal for low EA surveillance is not any single biomarker but a compound pattern. Here is what a “suggestive” compound pattern looks like, as a rough heuristic:

• HRV trending down more than 1 SD below baseline, sustained 2+ weeks
• RHR trending up more than 1 SD above baseline, sustained 2+ weeks
• Sleep duration or quality trending worse, 2+ weeks
• Body mass trending down during a period of consistent or increasing training
• Menstrual irregularity or missed period (if applicable)
• Subjective fatigue, blunted mood, or cold intolerance
• No obvious confounders active (no illness, no alcohol pattern, no travel, no major training spike, no luteal-phase timing effect that accounts for the autonomic signals)

When this pattern is present, an honest surveillance tool flags it as worth investigating — not as a diagnosis. The user’s next step is to consider confounders carefully and, if the pattern persists and is unexplained, speak with a sport dietitian or sport medicine physician.

Surveillance tools that flag on single-signal thresholds generate many more false positives than they do meaningful alerts. Tools that require the compound pattern are more conservative and more useful.

What Omnio’s Surveillance Actually Looks At

Omnio’s surveillance is feature-flagged and runs in shadow mode while the empirical false-positive rate is being evaluated. The design, as described in the pillar, looks for multi-signal rolling trends relative to personalized baselines, gated against cycle phase, illness, alcohol, training spikes, and travel. The output is a review-this-with-a-professional framing, never a diagnosis.

For the confounder-gating logic in detail, see confounders that mimic RED-S. For the EA calculation mechanics, see energy availability calculation explained. For what to do when a pattern is flagged, see refeed protocols.

Putting It Together

Low energy availability produces real changes in real biomarkers. HRV declines, RHR shifts, sleep fragments, body mass drops, menstrual function suppresses, strength adaptation blunts, mood dulls. None of these signals is unique to low EA. Each has at least three other common causes. The useful signal is the compound pattern, not the single biomarker.

Wearables can show you the signals. Clinical assessment is how the diagnosis happens. If your data is showing a compound pattern consistent with underfueling, bring the history to a sport medicine physician or sport dietitian. The data helps frame the conversation. It does not replace it.

Return to the energy availability and RED-S pillar for the full clinical framing. Related cluster reading: what is RED-S for the syndrome overview, confounders that mimic RED-S for the alternative explanations, and refeed protocols for response options. Cross-cluster: composite scores with confidence for how biomarker signals should be combined with explicit confidence.