Technology · Proxy DLMO
Dim-light melatonin onset,estimated without a laboratory.
Salivary DLMO under controlled dim light remains the clinical reference. Our free tier estimates the same phase marker from two population-validated behavioural signals — fused with explicit uncertainty bands.
Engine dlmo-proxy-v1
Reference vs proxy
Reference standard vs free-tier proxy
Salivary DLMO
Repeated samples under controlled dim light — clinical reference, high burden, not scalable at population entry.
Behavioural proxy
Sleep onset − 2 h and MCTQ mid-sleep − 2.5 h — published offsets, circular fusion, confidence capped at 0.55.
Proxy DLMO (free tier)
Dim-Light Melatonin Onset (DLMO) is when melatonin begins rising under dim light — the reference phase marker in chronotherapy. Lab DLMO needs repeated saliva samples. We estimate it from two published behavioural proxies, then fuse them:
- Sleep timing
DLMO ≈ habitual sleep onset − 2 h
Circular mean of recent phone or wearable sleep onsets. Habitual sleep onset typically follows DLMO by about two hours (Burgess et al., 2016).
- Chrono test (MCTQ)
DLMO ≈ mid-sleep on free days − 2.5 h
Mid-sleep corrected for sleep debt (MSFsc) from the Munich Chronotype Questionnaire — a population-validated phase marker (Roenneberg).
When both signals are present, we weight-fuse them: more synced nights increase weight on sleep timing; disagreement widens your uncertainty band (typically ±60–90 min). Confidence is capped well below clinical grade.
Algorithm
Four-step fusion
Live in estimateDlmoProxy() — same logic on every dashboard load.
- Extract
Circular mean sleep onset
Up to 14 nights in a 21-day window. Local wall-clock preserved — no server timezone conversion.
- Anchor
MCTQ MSFsc offset
Latest chronotype profile. DLMO ≈ sleep-corrected mid-sleep on free days − 2.5 h.
- Fuse
Weighted blend
Behavioural weight = min(14, nights); questionnaire weight = 2. Disagreement widens the uncertainty band.
- Report
Phase offset minutes
Deviation from population mean DLMO (21:00) — drives medication window shifts in the BTI engine.
Uncertainty
Confidence is capped — by design
A proxy never claims clinical certainty. Bands narrow as evidence accumulates.
- Low
±90 min
< 0.30 confidence
- Moderate
±75 min
0.30 – 0.44
- Strong proxy
±60 min
≥ 0.45 (max 0.55)
- Behavioural: +0.10 base, +0.03 per synced night (max +0.35).
- Questionnaire present: +0.12.
- Agreement ≤30 min: +0.10 · ≤60 min: +0.05 · >120 min: −0.05.
Upgrade path
Three-tier DLMO resolution
tiptraq_l1 dominates above 0.4 confidence; blood_panel_l2 above 0.3; else smartphone_l3.
- Free
Proxy DLMO from sleep
Phone and wearable sleep logs give habitual sleep onset. We apply DLMO ≈ sleep onset − 2 h (Burgess et al., 2016) — a behavioural proxy, not a lab measurement.
- Step 2
Chrono test refines phase
The Munich Chronotype Questionnaire (MCTQ) adds DLMO ≈ MSFsc − 2.5 h (Roenneberg). We fuse this with sleep timing when both are available.
- Clinical
TipTraQ validation
Three nights at home with the TipTraQ kit give a clinical-grade read — sleep staging, breathing, and oxygen. That block replaces the proxy and unlocks your verified badge.
What this is not
- Not salivary or plasma DLMO under controlled dim light.
- Wearable staging varies — we use onset timestamps, not device staging labels.
- No drug PK/PD simulation; windows are phase-adjusted chronotherapy offsets.
- Model weights stay server-side — UI receives timing payloads only.
TipTraQ three-night validation replaces the proxy and unlocks the verified clinical-grade badge.
Primary references
- 2016
Relationship between melatonin onset and sleep onset
Burgess, H.J. et al. · Sleep Medicine · Sleep onset ≈ 2 h after DLMO
- 2007
Epidemiology of the human circadian clock
Roenneberg, T. et al. · Sleep Medicine Reviews · MCTQ / MSFsc
Ready for clinical review?
Walk the triage panel or return to the full technology overview.