Where our numbers come from.

44 studies with 21,438 participants: RMSSD ≈ 42 ± 15 ms, SDNN ≈ 50 ± 16 ms over 5 minutes. RMSSD and SDNN are different metrics on different scales. Important for how this is read: the ± is the spread BETWEEN studies, not the distribution within a population — sound as a normative reference, not as an individual assessment.

HRV baselines and deviations

“Parasympathetic nerve activation contributes to at least 50 % of the LF variability while sympathetic activity, at best, only contributes 25 % to this variability.” That makes the LF/HF ratio unusable as a sympathovagal index — which is why Aellic deliberately does not use it.

HRV baselines and deviations

One night of total sleep deprivation (N = 13, randomised crossover): muscle protein synthesis −18 %, cortisol +21 %, testosterone −24 %. Acute anabolic resistance — the basis for how heavily sleep weighs in the recovery score.

Recovery score weighting and age references

13 healthy adults, 536 nights against a Polar H10 chest strap (single-lead ECG) as the reference. Resting heart rate reached Lin's concordance correlation of 0.86 (Polar Grit X Pro, which the authors themselves grade “poor”) to 0.98 (Oura Gen 4), with mean absolute percentage errors of 1.67–3.00 %. The corresponding HRV errors were three to five times larger on the same nights and devices (MAPE 5.96–16.32 %). That error ratio — not the concordance coefficient — is what supports treating nocturnal resting heart rate as the steadier of the two signals.

Recovery score weighting and age references

ln(RMSSD) is the most practical measure for day-to-day monitoring; the log transform normalises the right-skewed RMSSD distribution, and the paper establishes that averaging is required at all, rather than reading single measurements. NOT supported are the specific window lengths (14 nights minimum, 7 days acute, 30 days chronic) — those are design decisions.

HRV baselines and deviations

24-hour norms for 260 healthy people (112 m / 148 f) aged 10 to 99, by decade. The actual primary source for the age dependence of HRV, which Aellic previously misattributed to Shaffer & Ginsberg. The decline is NOT linear: RMSSD falls to 47 % of its starting value by the sixth decade (pNN50 to 24 %) “and then stabilized”. A constant rate of “ms per year of life” cannot be derived from it.

HRV baselines and deviations

Age-dependent sleep duration recommendations from an expert panel: 7–9 h for young adults and adults, 7–8 h from age 65. A population default, sound as a cold start, not as an individual target. Full report: Sleep Health 1(4):233–243, doi 10.1016/j.sleh.2015.10.004.

Recovery score weighting and age references

There is no reliable single biomarker for overtraining: “none of them meet all the criteria to make their use generally accepted”. A multi-signal approach beats any individual marker — hence the corroboration through resting heart rate and respiratory rate. The statement names NO threshold values, however; the specific limits (+5 bpm, +1.5 brpm) are Aellic's own. Second publication: European Journal of Sport Science 13(1):1–24, doi 10.1080/17461391.2012.730061 — different pagination, do not mix the two.

HRV baselines and deviations

The reference definition of the time- and frequency-domain HRV measures. Basis for the separate RMSSD and SDNN tables. Deliberately recorded WITHOUT a PMID: the frequently co-cited PMID 8737210 belongs to the second publication in the European Heart Journal 17(3):354–381 (doi 10.1093/oxfordjournals.eurheartj.a014868), not to this Circulation version — the combination would be wrong in itself.

HRV baselines and deviations

24 studies: caffeine reduced total sleep time by 45 min and sleep efficiency by 7 %, lengthened sleep onset latency by 9 min and wake after sleep onset by 12 min, and reduced deep sleep by 11.4 min. To avoid a loss of total sleep time, a 107 mg coffee should be taken at least 8.8 h before bedtime and a 217.5 mg pre-workout serve at least 13.2 h before.

What shapes your nights: days compared with the night after

Meta-analysis across 65 studies / 3,577 people aged 5 to 102. WASO is the largest age effect in sleep architecture — rising by roughly 10 min per decade of life between 30 and 60, largely stable thereafter. The peer-reviewed basis for normative bands of sleep architecture; the N3 and REM bands used in Aellic have not yet been reconciled line by line against this paper's tables.

Sleep stages and bedtime

The two-process model: sleep propensity is the interaction of homeostatic sleep pressure (Process S, rising exponentially during wake and falling exponentially during sleep) and a circadian rhythm (Process C). Its consequence for any sleep-debt feature: Process S SATURATES — sleep debt is not a bank account, it neither accumulates linearly forever nor is it repaid 1:1. The model supports only the qualitative shape (bounded, recency-weighted); every concrete window, decay or cap constant is construction, not literature.

Bedtime plan from wake time, sleep need and sleep debt, Sleep stages and bedtime

One night of total sleep deprivation (N = 13, randomised crossover): muscle protein synthesis −18 %, cortisol +21 %, testosterone −24 %. Acute anabolic resistance — the basis for how heavily sleep weighs in the recovery score.

Sleep stages and bedtime

n = 35 against laboratory polysomnography: Oura performs best among the consumer devices (κ 0.65 on the four-stage classification, ahead of Apple Watch 0.60 and Fitbit Sense 0.55) — but remains far from polysomnography.

Sleep stages and bedtime

UK Biobank (n = 60,977, 1,859 deaths): sleep regularity is a STRONGER predictor of mortality than sleep duration — 20–48 % lower all-cause mortality in the upper four SRI quintiles compared with the lowest.

Sleep stages and bedtime

n = 88,975: hazard ratio 1.53 (95 % CI 1.41–1.66) at the 5th SRI percentile.

Sleep regularity

n = 53 across 6 devices against polysomnography: agreement on multi-class sleep stage only 50–65 %, κ 0.20–0.52 (Apple 0.20 up to Somfit 0.52). Sleep sensitivity 90–98 %. This is why the stage distribution now weighs only 0.10 in Aellic instead of 0.25.

Sleep stages and bedtime

Definition of the Sleep Regularity Index: SRI = 100 × (2 × P(same state at t and t+24 h) − 1), on a scale of −100 to 100. The paper does not ask for “at least 7 days” but for whole MULTIPLES of 7 days — a weekly grid, so that the distribution of weekdays cannot skew the result.

Sleep regularity, Sleep stages and bedtime

Definition of social jetlag: SJL = |MSF − MSW|, the difference in mid-sleep between free days and work days (n = 501, Munich ChronoType Questionnaire). Limit of our verification: the full text is paywalled; the formula itself was confirmed only through the citing literature, not read in the original.

Sleep regularity

Age-dependent sleep duration recommendations from an expert panel: 7–9 h for young adults and adults, 7–8 h from age 65. A population default, sound as a cold start, not as an individual target. Full report: Sleep Health 1(4):233–243, doi 10.1016/j.sleh.2015.10.004.

Bedtime plan from wake time, sleep need and sleep debt

Consensus by modified RAND Appropriateness Method: adults should sleep 7 or more hours per night on a regular basis to promote optimal health. Sleeping more than 9 hours may be appropriate for some groups; the statement sets a floor, not an optimum.

Bedtime plan from wake time, sleep need and sleep debt

Network meta-analysis of wearable-based ovulation detection: exact-day accuracy pooled across devices is 0.56, and only ~68 % land within ±1 day. This is the justification for Aellic ALWAYS shipping the estimated ovulation with a confidence band and never presenting it as “measured”.

Phases and cycle-aware baselines

78 studies, n = 1,193: ES −0.04, CrI [−0.11, 0.08], GRADE low. There is NO appreciable luteal performance decrement. The popular advice to “take it easier in the luteal phase” inverts a finding that is trivial to begin with — which is why Aellic builds no cycle-phase-driven training recommendation.

Phases and cycle-aware baselines

37 studies: meta-analytic d = −0.39, 95 % CI [−0.67, −0.11]; the primary literature is “not only inconsistent but contradictory”, two studies found the opposite direction, and only 42 % verified the phase hormonally. Too weak for a fixed constant — hence no cold-start offset for HRV.

Phases and cycle-aware baselines

Umbrella review of meta-analyses and systematic reviews on cycle-phase-driven training: “it is premature to conclude that short-term fluctuations in reproductive hormones appreciably influence acute exercise performance or longer-term strength or hypertrophic adaptations”.

Phases and cycle-aware baselines

42,759 cycles from 2,596 women across 1.2 million days: the amplitude of the cyclical swing scales with cycle length; measured across respiratory rate, resting heart rate, HRV, temperature and SpO₂. Preprint of the same work: bioRxiv, doi 10.1101/2025.09.11.675620 — the peer-reviewed version is deliberately the one cited here.

Phases and cycle-aware baselines

Validation of ovulation detection through a temperature wearable; confirms the order of magnitude of MAE 1.2–1.3 days from wang2025 in a second data set and on a second device.

Phases and cycle-aware baselines

Retrospective ovulation estimation from wrist temperature: MAE 1.2–1.3 days, 89 % of estimates within ±2 days (PAE2). Carries three constants of the temperature-based ovulation estimate: the signal threshold of +0.2 °C, the confidence band of ±2 days, and the honesty rule that a non-detection is not a finding — only ~80.8 % of cycles reach the threshold at all, and ~19 % produce no usable signal whatsoever. A non-detection therefore explicitly does NOT mean “anovulatory”.

Phases and cycle-aware baselines

n = 11,590: only 80.6 % of women show the expected RMSSD amplitude across the cycle. A fixed offset would impose a pattern on one woman in five that she does not have — THE load-bearing argument for a per-person learned offset instead of a constant from the literature. It also supplies the shape that the two-strata model deliberately simplifies: resting heart rate has its minimum on cycle day 5 and its maximum only on day 26, so it rises during the fertile window already, rather than jumping after ovulation.

Phases and cycle-aware baselines

Users of combined oral contraceptives are NOT signal-free: nocturnal HRV follows a real rhythm driven by the pill pack (higher during the pill-free interval). Aellic's strata sit at estimated ovulation, not at the pack boundary — so part of this is captured, but not all of it. A genuine pill-pack mode would be a development stage of its own; until then the position is: partially correct, never wrong, and documented.

Phases and cycle-aware baselines

Wrist skin temperature against oral basal temperature for ovulation detection; supplies the upper end of the luteal temperature rise (up to +0.50 °C) and confirms that skin carries the larger amplitude. Aellic deliberately does NOT take this value but the lower edge from maijala2019 — too large an offset would mask genuine strain.

Phases and cycle-aware baselines

Progesterone, not estradiol, predicts the lower luteal HRV. That makes it an endocrine signal, not exhaustion — penalising it as a recovery deficit was a category error, and that is the justification for the entire seasonal adjustment.

Phases and cycle-aware baselines

612,613 cycles with ovulation confirmed by basal temperature and LH: cycle length 29.3 ± 5.2 d, follicular phase 16.9 ± 5.3 d, luteal phase 12.4 ± 2.4 d. Variance ratio follicular:luteal ≈ 4.9:1 — all of the spread sits BEFORE ovulation. This is why ovulation is estimated backwards from the next expected period: going backwards inherits the luteal SD of 2.4 instead of the follicular one of 5.3. It also carries `DEFAULT_CYCLE_LENGTH = 29` (only 13 % of all cycles really run 28 days; the mean is 29.3).

Phases and cycle-aware baselines

708 cycles: resting heart rate +3.96 bpm from the follicular to the late luteal phase. The luteal RESPIRATORY RATE swing, by contrast, is only +0.2 to +0.6 br/min and therefore BELOW the night-to-night precision of commercial sensors (~1.0 br/min) — which is why there is deliberately no cold-start offset for `respiratory_rate`, only the learned one.

Phases and cycle-aware baselines

Nocturnal FINGER skin temperature across the cycle, measured with an Oura ring: luteal rise +0.30 ± 0.12 °C. That is the lower edge of the range Aellic uses and therefore the cold-start value for `body_temp`. Important for how this is read: skin, not core — distal skin temperature has the LARGER amplitude compared with an oral basal temperature (~0.20–0.23 °C), and skin is exactly what a wearable measures.

Phases and cycle-aware baselines

2.7 million cycles: only ~24 % of ovulations fall on cycle day 14–15, and 90 % are spread across days 10–24. Anyone who simply assumes “day 14” is wrong in a good three cases out of four.

Phases and cycle-aware baselines

Wrist skin temperature: luteal rise +0.33 °C; a sustained 3-day shift occurs in 82 % of cycles, and 86 % of shifts begin on the day of ovulation or later. Carries three constants of the ovulation estimate at once: the 3-day rule (`TEMP_SHIFT_SUSTAIN_DAYS`), the choice of the day BEFORE the shift begins as ovulation, and the stratum boundary `> ovu_day`.

Phases and cycle-aware baselines

274 cycles: pulse rate during sleep rises mid-luteally by +3.8 bpm compared with menstruation. One of three independent data sets behind the range of +2 to +4 bpm (the others: goodale2019 +3.96, jasinski2024 ~3.5 peak-to-trough). Aellic sets the cold start to +2.0 — the lower edge.

Phases and cycle-aware baselines

Progesterone acts as a respiratory drive; minute ventilation rises luteally by about 1 L/min. The effect is real — but it runs almost entirely through TIDAL VOLUME, not through rate. Which is why, despite demonstrable physiology, all that remains in respiratory rate is a swing below device precision (see goodale2019), and why there is no cold-start constant for it. Listed in PubMed without a DOI.

Phases and cycle-aware baselines

Comparison of 24-hour body temperature between men, naturally cycling women and women using hormonal contraceptives. Establishes the honest limit of the cycle engine: on combined pills the temperature stays chronically raised and flat — it does not even fall back during the pill-free interval. The learned temperature offset therefore correctly tends towards zero there. Listed in PubMed without a DOI. The addition “n = 8, rectal” found in the code prose has not been verified against the full text.

Phases and cycle-aware baselines

Luteal length 14.13 ± 1.41 d — measured from the LH PEAK. Aellic has no LH test and must therefore use bull2019 (12.4 d from estimated ovulation); mixing the two anchors would shift every phase boundary by ~2 days. This anchor trap is exactly why `LUTEAL_LENGTH_DAYS = 12` rather than the reflexive 14. No DOI: published in 1984, before this journal assigned them.

Phases and cycle-aware baselines

Meta-regression: “a 0.24 % increase in hypertrophy (95 % CrI: 0.15, 0.33) per additional set at the average ‘fractional’ weekly volume of 12.25 sets”.

Muscle growth — sets per muscle

15 studies. Training to set failure held a “trivial advantage” over non-failure for hypertrophy (ES 0.19; 95 % CI 0.00–0.37; p = 0.045), and none at all for momentary muscular failure specifically (ES 0.12; p = 0.343). Proximity was graded by velocity-loss thresholds, where high (>25 %) versus moderate (20–25 %) also came out level (ES 0.08; p = 0.529). Supports the direction — closer to failure is not worse — and nothing more precise than that.

Muscle growth — sets per muscle

15 studies in young adults. No significant difference between failure and non-failure training, for strength (ES −0.09; 95 % CI −0.22 to 0.05) or for hypertrophy (ES 0.22; 95 % CI −0.11 to 0.55). One subgroup runs the other way and matters for Aellic's audience: among already resistance-trained participants, training to failure WAS significantly better for hypertrophy (ES 0.15; 95 % CI 0.03–0.26).

Muscle growth — sets per muscle

The fibre-level follow-up to schumann2022, and the one place where a small interference effect does show up: 15 studies, SMD −0.23 for overall muscle fibre hypertrophy (95 % CI −0.46 to −0.00; p = 0.050), −0.34 for type I (p = 0.078) and −0.13 for type II (p = 0.315). This is also the source for the running-versus-cycling difference: type I fibres were negatively affected when the aerobic work was running (SMD −0.81; 95 % CI −1.26 to −0.36) but not cycling. The authors call the contrast with whole-muscle findings “intriguing” rather than resolved.

Muscle growth — sets per muscle

43 studies, 1,090 participants — the whole-muscle picture, and the reason Aellic does not warn about interference. Hypertrophy SMD −0.01 (p = 0.919) and maximal strength −0.06 (p = 0.446) are untouched; only explosive strength is attenuated (−0.28; p = 0.007), and more so when both sessions fall in the same session than when they are separated by at least three hours (p = 0.043). Type of aerobic training, training frequency, training status and age were all non-significant moderators.

Muscle growth — sets per muscle

With weekly volume equated, training frequency has no meaningful influence on hypertrophy. THIS is the source for Aellic's decision not to score frequency separately — not the 2016 predecessor, which was wrongly cited for it until 2026-07-25.

Muscle growth — sets per muscle

34 treatment groups from 15 studies: “Each additional set was associated with an increase in effect size (ES) of 0.023 corresponding to an increase in the percentage gain by 0.37%” (p = 0.002). The dose is coded as total sets per muscle group per week.

Muscle growth — sets per muscle

HRmax = 208 − 0.7 × age. Meta-analysis across 351 studies / 492 groups / 18,712 people (r = −0.90), cross-validated in a laboratory cohort of 514 healthy people whose own regression, 209 − 0.7 × age, the authors call “virtually identical”. Individual values scatter about ±10 bpm around the line, so the formula sets a population expectation and not a personal ceiling. The widespread 220 − age formula has no comparable derivation.

Daily strain: session load, recommended range, weekly corridor, Training load — zones, TRIMP, ACWR

Six studies in trained men aged 18–35, binned a priori into “low” (<12 weekly sets), “moderate” (12–20) and “high” (>20). Moderate and high volumes did not differ for the quadriceps (p = 0.19) or the biceps brachii (p = 0.59); for the triceps brachii the HIGHER volume was significantly better (p = 0.01). The authors' recommendation is “a range of 12–20 weekly sets per muscle group may be an optimum standard recommendation”, offered as pragmatic advice for that population.

Muscle growth — sets per muscle

EMG in the deadlift: erector spinae and quadriceps show higher activation than gluteus maximus and biceps femoris. Confirms the mapping to the erector spinae. For the Romanian variant the relationship reverses.

Which muscles an exercise loads

Severe methodological criticism of the ACWR: “there are known issues with the use of ratio data”, it is “an inaccurate metric (failing to normalize the numerator by the denominator even when uncoupled)”, “not consistently and unidirectionally related to injury risk”, and “there is no evidence supporting the use of ACWR in training-load-management systems”. This is why Aellic reports the ACWR only with the label “heuristic”.

Daily strain: session load, recommended range, weekly corridor, Training load — zones, TRIMP, ACWR

Acute EMG amplitude is NOT a validated predictor of hypertrophy; conclusions about long-term adaptation are “frequently unsubstantiated and unwarranted”. The muscle mapping may therefore only say which muscles a movement loads — not which one grows most. Note: year 2018 with volume 8 and a 2017 DOI stem is how this was actually published (eCollection 2017), not a typo.

Muscle growth — sets per muscle, Which muscles an exercise loads

The actual origin of the four-week chronic window Aellic computes against: injury likelihood is modelled against “acute load spikes above what they have been doing on average over the past 4 weeks (chronic load)”, with a polynomial fit of R² = 0.53 across three sports. Also the source of the sweet-spot figure that gabbett2016 reproduces.

Daily strain: session load, recommended range, weekly corridor, Training load — zones, TRIMP, ACWR

Source of the corridor Aellic ships, twice stated verbatim: “acute:chronic workload ratios within the range of 0.8–1.3 could be considered the training ‘sweet spot’, while acute:chronic workload ratios ≥1.5 represent the ‘danger zone’”. Acute load is one week of training, “a logical and convenient unit”.

Daily strain: session load, recommended range, weekly corridor, Training load — zones, TRIMP, ACWR

Strength scales with body mass^(2/3) under geometric similarity. Linear division systematically penalises heavier athletes. Important limitation: the exponent 0.67 applies to muscle STRENGTH (dynamometer); for torque (isokinetic) the paper explicitly recommends b = 1, so linear division is the correct choice there. Scaling a torque-derived measure by ^(2/3) would be a misapplication of this source.

Strength standards and ranks

TRIMP weights intensity exponentially (lactate-weighted) and computes on heart rate reserve — not in linear steps on raw %HRmax. A book chapter, hence no DOI. Two caveats: the chapter's content is confirmed only through the citing literature, not read in the original; and the frequently cited book title “Physiological Testing of Elite Athletes” belongs to the first edition of 1982 — the source here is the second edition of 1991.

Training load — zones, TRIMP, ACWR

25 untrained men, 8 weeks, unilateral leg extension, volume equated. At 30 % 1RM failure IS required (quadriceps CSA +7.8 % to failure versus +2.8 % not to failure); at 80 % it is not (+8.1 % versus +7.7 %). Strength followed load rather than failure (~33 % at 80 % 1RM versus ~17 % at 30 %).

Muscle growth — sets per muscle

With volume load equated, 40, 60 and 80 % 1RM were equivalent, “however, 20 % 1RM was suboptimal” (vastus lateralis cross-sectional area +8.9 % at 20 % versus ~20 % for all the others).

Muscle growth — sets per muscle

Confirms Mitchell 2012 in already-trained men over 12 weeks (~30–50 % versus ~75–90 % 1RM, all sets to failure): load is secondary for hypertrophy, proximity to failure is what counts.

Muscle growth — sets per muscle

Hypertrophy is comparable across 30–80 % 1RM provided the work is taken close to muscular failure (muscle volume by MRI +6.8 % at 30 %, +7.2 % at 80 %, p = 0.18). Only the two end points were tested, no intermediate loads.

Muscle growth — sets per muscle

809,986 drug-tested raw competition entries (571,650 male, 238,336 female). For male lifters aged 18–35 the 90th percentile lands almost exactly on Aellic's elite anchor: squat 2.83 versus our 2.75, bench press 1.95 versus 2.00, deadlift 3.25 versus 3.25 (all as 1RM ÷ body mass). Peer-reviewed backing for three of the twelve lifts, on the male side. The female elite anchors rest on strengthlevel alone; this paper's female 90th percentiles are lower (squat 2.26, bench press 1.35, deadlift 2.66) and were never used to set them.

Strength standards and ranks

Overhead pressing involves the upper trapezius across all four variants tested (its task: stabilising scapular elevation) — which justifies adding `traps` to the vertical push family. Limitation: the paper publishes effect sizes only, no absolute %MVIC values; HOW strongly the upper trapezius contributes cannot be quantified from it.

Which muscles an exercise loads

EMG in the Romanian deadlift: biceps femoris is the most active of the muscles measured (56.66 % peak versus gluteus maximus 46.88 % and rectus femoris 25.26 %). Confirms the hamstring mapping.

Which muscles an exercise loads

EMG of the hip thrust versus the squat: upper gluteus maximus 69.5 % versus 29.4 %, lower 86.8 % versus 45.4 %. Confirms the gluteus mapping.

Which muscles an exercise loads

Why a repetition-based estimate degrades at higher rep counts for reasons no formula can fix: at 70 % 1RM endurance runners managed 39.9 ± 17.6 repetitions against weightlifters' 17.9 ± 2.8, and at 80 % 19.8 ± 6.4 against 11.8 ± 2.7 (both p < 0.05); at 90 % the difference vanished. Above roughly 10 repetitions the estimate depends more on training history than on which equation is used. Small sample — 8 athletes per group.

Workout logging — 1RM estimation

Directly measured VO2peak by decade in 3,816 healthy Norwegian adults. Men: 54.4 (20s), 49.1 (30s), 47.2 (40s), 42.6 (50s), 39.2 (60s), 35.3 (70+); women: 43.0, 40.0, 38.4, 34.4, 31.1, 28.3 ml/kg/min, each with an SD of about 5 to 8. Decline about 3.5 ml/kg/min per decade. Reading the fitness age as the age whose decade mean equals the user's estimated VO2max, interpolated between decade midpoints, is Aellic's construction on top of this table, not a result of the paper.

Fitness age and VO2max estimate

70 subjects aged 18–69, bench press and leg press at 1, 5, 10 and 20RM. The load-repetition relationship is curvilinear, not linear (bench press linear Sy.x = 2.6 kg versus non-linear 0.2 kg; leg press 11.0 versus 2.6 kg), and accuracy falls as the repetition count rises (R² for 5/10/20RM: bench 0.993/0.976/0.955, leg press 0.974/0.933/0.915). Concludes verbatim: “no more than 10 repetitions should be used in linear equations to estimate 1RM”. This is the source of Aellic's 10-repetition cap.

Workout logging — 1RM estimation

VO2max can be estimated as 15.3 x HRmax / HRrest. Derived in 46 well-trained men aged 21 to 51, so the sample is small, male and fit; for everyone else the ratio gives a rough population-level estimate, labelled as estimated in the app. Aellic applies it only when no device-measured VO2max exists in the last 60 days.

Fitness age and VO2max estimate

sRPE = RPE (0–10) × duration in minutes, deliberately collected only 30 minutes after the session, “so that particularly difficult or particularly easy segments toward the end of the exercise bout would not dominate the subject's rating”. Works entirely without a heart rate signal and also yields monotony and strain (Table 5).

Training load — zones, TRIMP, ACWR

In water the heart rate runs lower than on land: 190 versus 175 bpm maximal (15 bpm) and 165 versus 152 bpm at the ventilatory threshold (13 bpm). Without an offset, every HR zone systematically overestimates intensity in water.

Training load — zones, TRIMP, ACWR

The paper from which the heart rate reserve method emerged: personalise zones through the reserve, that is, take resting heart rate into account rather than maximum heart rate alone. Assessment: the closed formula THR = HRrest + f × (HRmax − HRrest) is a later textbook formalisation, not a formula this small longitudinal study puts forward itself. PubMed carries no abstract; the content is evidenced only through secondary literature.

Daily strain: session load, recommended range, weekly corridor, Training load — zones, TRIMP, ACWR

EMG comparison of eight back exercises in 19 trained participants: rowing movements activate the middle trapezius most strongly, pull-up and chin-up the latissimus.

Which muscles an exercise loads

Target / synergist / stabiliser framework, translated into primary and secondary. Convention, not measurement — the mapping was cross-checked against EMG for each movement pattern.

Muscle growth — sets per muscle

Percentile bands from roughly 195 million reported lifts (beginner 5th to elite 95th percentile). Crowd-sourced self-report with an upward bias — for 9 of the 12 lifts it is the only source, and it is labelled as exactly that. For the male squat, bench press and deadlift anchors vandenhoek2024 provides peer-reviewed agreement; the female anchors rest on this source alone.

Strength standards and ranks

1RM estimate w × (1 + reps/30). Grey literature, and worth stating plainly: the source is a self-published training chart, not a study — the formula is an algebraic reconstruction of that table, with no sample and no derivation data behind it. Richens & Cleather 2014 describe it in their own comparison table as “a poundage chart, not based on scientific research”. Aellic uses it because it is the established convention and performs acceptably in validation work, not because it is well founded. The cap at 10 repetitions is carried by reynolds2006.

Strength standards and ranks, Workout logging — 1RM estimation

WHO statement: no level of alcohol consumption can be established as safe for health, and ethanol is classified by IARC as a group 1 carcinogen. Carries the decision to downgrade any drink above 1.2 % ABV in the food score, independently of its energy and sugar.

Food scoring

Thirteen beverages measured against still water as the reference. Cumulative urine output over 4 h after cola, diet cola, hot tea, iced tea, coffee, lager, orange juice, sparkling water and a sports drink was not different from water. Only an oral rehydration solution, full-fat milk and skimmed milk retained more, with a mean beverage hydration index at 2 h of 1.54, 1.50 and 1.58 respectively. This is why every non-alcoholic drink in Aellic counts in full, and why beer does too.

How drinks count towards your fluid target

Fifty male habitual coffee drinkers, counterbalanced cross-over over three days: 4 x 200 ml of coffee providing 4 mg/kg caffeine per day against the same volume of water. No difference in total body water, in any haematological marker, or in 24 h urine volume. The authors conclude that coffee in moderation provides similar hydrating qualities to water. Carries the decision to count coffee and tea at a factor of 1.0 instead of the 0.85 that fitness apps commonly apply without a source.

How drinks count towards your fluid target

Twelve males drank 1 l of beer the morning after being dehydrated by 1.9 % of body mass, twice rehydrated beforehand and twice not. With 4 % ethanol against the same alcohol-free beer, urine output differed when euhydrated (1279 vs 1121 ml) but not when hypohydrated (261 vs 174 ml). The authors conclude that the diuretic action of alcohol is blunted when the body is already hypohydrated.

How drinks count towards your fluid target

The published Nutri-Score algorithm in its current version. The thresholds and the fruit and vegetable bonus of the Aellic Health Score follow this specification. Note that there are TWO documents — foods adopted on 2022-06-29, beverages on 2023-02-01 (beverage report: https://www.santepubliquefrance.fr/sites/default/files/rdd/document/Update%20report%20beverages_31%2001%202023-VOTED.pdf).

Food scoring

Nine studies with 480,479 participants: stair climbing was associated with lower cardiovascular mortality (RR 0.61, 95 % CI 0.48–0.79) and all-cause mortality (RR 0.76, 95 % CI 0.62–0.94). Pooled from observational data; no causal claim and no dose, such as flights per day, can be drawn from it.

Suggested routines

57 studies from 35 cohorts: non-linear dose-response curves with inflection points around 5,000–7,000 steps a day. Compared with 2,000 steps, 7,000 steps was associated with 47 % lower all-cause mortality (HR 0.53). Evidence certainty mostly moderate; observational cohorts with possible residual confounding. An erratum appeared in Lancet Public Health 10(9):e731.

Suggested routines

27 studies: alcohol delayed REM onset and shortened REM duration dose-dependently, starting at a low dose (up to 0.50 g/kg, roughly two standard drinks). Shorter sleep onset latency appeared only at high doses (0.85 g/kg or more, roughly five drinks); effects on total sleep time, sleep efficiency and wake after sleep onset remained uncertain.

Suggested routines

Preregistered meta-analysis of 10 studies (n = 4,674, 38 effect sizes): complete social media abstinence had no significant effect on positive affect, negative affect or life satisfaction, and abstinence duration did not moderate this. Addresses abstaining, not limiting use; methodological differences between studies are large.

Suggested routines

77 studies, 186 effect sizes: stretch training for two weeks or more increased joint range of motion with a large effect versus controls (ES about 1.0). Static stretching and PNF outperformed ballistic or dynamic stretching; volume, intensity and frequency were not significant moderators. Range of motion only; it says nothing about injury or pain.

Suggested routines

Systematic review with six meta-analyses (k = 56, n = 7,093): only working memory capacity showed a significant negative effect of phone presence, smaller than theorised; all other cognitive functions had null pooled effects. Substantial heterogeneity and generally low power in the field. Published online 2023-11-23.

Suggested routines

24 studies: caffeine reduced total sleep time by 45 min and sleep efficiency by 7 %, lengthened sleep onset latency by 9 min and wake after sleep onset by 12 min, and reduced deep sleep by 11.4 min. To avoid a loss of total sleep time, a 107 mg coffee should be taken at least 8.8 h before bedtime and a 217.5 mg pre-workout serve at least 13.2 h before.

Suggested routines

Seven single-day randomised crossover trials, mostly in overweight adults: frequent light-intensity walking breaks lowered postprandial glucose compared with continued sitting (d = −0.72), and more than standing breaks did (d = −0.30). Standing alone had a smaller effect (d = −0.31). No effect on systolic blood pressure. Acute laboratory effects only.

Suggested routines

16 cohort studies: muscle-strengthening activity was associated with 10–17 % lower risk of all-cause mortality, cardiovascular disease, total cancer and diabetes, independent of aerobic activity. The associations were J-shaped, with the largest risk reduction at about 30–60 min a week; the effect of higher volumes is unclear.

Suggested routines

Participant-level data from 15 cohorts (47,471 adults, 3,013 deaths): compared with the lowest quartile (median 3,553 steps a day), the third quartile (median 7,842 steps) had an adjusted hazard ratio of 0.55. Risk fell progressively up to 6,000–8,000 steps a day in adults aged 60 or over, and up to 8,000–10,000 in younger adults. Observational.

Suggested routines

About 5 servings a day (2 fruit, 3 vegetables) was associated with the lowest mortality, with no further reduction above that (5 versus 2 servings: RR 0.87 across 26 cohorts, 1.89 million participants). Fruit juice and potatoes were not associated with mortality. The basis for the number 5 and for not counting juice.

Suggested routines

37 trials in the meta-analysis: activity breaks moderately attenuated postprandial glucose (SMD −0.54), insulin (SMD −0.56) and triacylglycerol (SMD −0.26) compared with uninterrupted sitting, with larger glucose effects at higher BMI. Supports the principle of breaking up sitting; it does not establish an hourly interval.

Suggested routines

185 prospective studies and 58 trials: comparing the highest with the lowest fibre intakes, observational data show 15–30 % lower incidence of several non-communicable diseases, with the greatest risk reduction at 25–29 g fibre/day. Similar findings for whole grains. GRADE certainty: moderate for fibre, low to moderate for whole grains.

Suggested routines

49 RCTs, 1,863 participants, at least 6 weeks of resistance training: protein supplementation added 0.30 kg fat-free mass and 2.49 kg to 1RM; no further gain in fat-free mass above a total intake of about 1.62 g/kg/day. It concerns total daily intake with resistance training and says nothing about per-meal distribution. Erratum: Br J Sports Med 2020;54(19):e7.

Suggested routines

95 prospective studies. Per 200 g/day of fruit and vegetables: RR 0.92 for coronary heart disease, 0.84 for stroke, 0.97 for total cancer and 0.90 for all-cause mortality; risk reductions were observed up to 800 g/day (600 g/day for cancer). Observational data.

Suggested routines

Gratitude interventions improved psychological well-being compared with measurement-only controls (d = 0.31) and with alternative activities (d = 0.17), but did no better than psychologically active comparisons (d = −0.03); no effect on anxiety. The effects are small.

Suggested routines

47 trials with active controls, 3,515 participants: moderate evidence that mindfulness programmes improve anxiety (ES 0.38 at 8 weeks), depression (0.30) and pain (0.33); low evidence for stress/distress; low or insufficient evidence for positive mood, attention, sleep and weight. Not better than active treatments. The trials tested structured multi-week programmes, not a daily number of minutes.

Suggested routines

32 studies. In adult cohorts, each additional daily serving of sugar-sweetened beverages was associated with 0.12–0.22 kg extra weight gain per year; adult RCTs that added these beverages increased body weight by 0.85 kg.

Suggested routines

148 studies, 308,849 participants: stronger social relationships went with 50 % higher odds of survival (OR 1.50), and 1.91 for complex measures of social integration. Observational; it says nothing about how often contact should happen.

Suggested routines

11 prospective cohorts, 310,819 participants: the highest intake of sugar-sweetened beverages (mostly 1–2 servings/day), compared with the lowest, was associated with RR 1.26 for type 2 diabetes and 1.20 for metabolic syndrome.

Suggested routines

15 studies in healthy people breathing below 10 breaths/min: higher HRV and respiratory sinus arrhythmia, more EEG alpha power, and more relaxation and comfort with less arousal. Mostly small, mechanistic studies; no pooled estimate and no dose.

Suggested routines

22 studies, 15 of them on dietary self-monitoring: a significant association between self-monitoring and weight loss was found consistently, but the authors grade the level of evidence as weak (mostly white, female samples; reliance on self-report). It supports logging intake, not any specific calorie target.

Suggested routines

Narrative review of 104 studies found through several databases: many physiological effects, but conclusions are limited by small, often single-sex samples and differing water temperatures. The authors call the health effects unresolved.

Suggested routines

Narrative review: with bedding and clothing as used in real life, heat exposure increases wakefulness and reduces slow-wave and REM sleep, and humid heat adds to the thermal load; cold exposure did not alter sleep stages but did affect cardiac autonomic responses. It names no optimal bedroom temperature.

Suggested routines

Control of the false discovery rate across a family of tests. Necessary because the insight engine computes ~10 habits × ~12 metrics ≈ 120 tests — of which roughly 6 would break the p < 0.05 threshold by chance alone and be shown to the user as advice. A note on grading: `strength = review` is a makeshift here — this is an original methodological work with proof and simulation, for which the registry's vocabulary holds no value of its own. No PMID; the paper predates PubMed coverage of this journal.

Statistical safeguards

Month-long randomised trial: blocking all mobile internet on participants' own phones for two weeks, with calls, texts and desktop internet still available and compliance tracked by app, improved subjective well-being, mental health and objectively measured sustained attention. Partly mediated by more time spent socialising in person, exercising and outdoors.

Suggested routines

111 healthy students (baseline 276 min a day), non-blinded: three weeks limited to 2 h a day or less produced small-to-medium improvements in well-being, stress, sleep quality and depressive symptoms versus control. Screen time rebounded after the intervention, approaching baseline by the six-week follow-up.

Suggested routines

Remote RCT: 5 minutes a day for one month. Breathwork, especially exhale-focused cyclic sighing, improved mood more and lowered respiratory rate more than mindfulness meditation did (p < 0.05). The senior author became a WHOOP advisor in June 2022.

Suggested routines

20 healthy adults, randomised parallel groups (n = 10 each): one night of sleep under 100 lx overhead room light, compared with less than 3 lx, raised night-time heart rate, lowered HRV and increased next-morning insulin resistance. A single night in a small sample, testing room light rather than a small standby LED.

Suggested routines

Preregistered direct replication of Ward 2017, Experiment 2, with six conditions (phone on desk, in pocket or bag, or outside the room; powered on or off): no difference between phone locations on the operation span task or the go/no-go task. The brain drain effect did not replicate.

Suggested routines

Randomised pilot, n = 38: avoiding the phone for 30 min before bed over four weeks reduced sleep latency and pre-sleep arousal and increased sleep duration and self-rated sleep quality compared with no instruction. Small, unblinded, with self-reported outcomes, but it tests exactly the 30-minute dose.

Suggested routines

Randomised field experiment (n = 237): notifications delivered as usual, batched three times a day, or never. The batched group reported feeling more attentive, in a better mood and in greater control of their phones, with lower stress and fewer interruptions. Receiving no notifications at all brought few of those benefits and higher anxiety and fear of missing out. A single, self-reported trial that has not been replicated.

Suggested routines

143 undergraduates randomised to limit Facebook, Instagram and Snapchat to 10 min per platform per day, about 30 min in total, or to use them as usual for three weeks: the limited group showed larger reductions in loneliness and depressive symptoms. Both groups reported less anxiety and fear of missing out, pointing to a self-monitoring effect. One small student sample.

Suggested routines

57 young adults, 5 minutes of writing at bedtime, sleep measured by polysomnography: people who wrote a to-do list fell asleep significantly faster than those who wrote about activities they had completed; the more specific the list, the faster they fell asleep. There was no no-writing control group.

Suggested routines

Two experiments proposing that the mere presence of one's own smartphone reduces available cognitive capacity, most strongly in highly dependent users. A preregistered direct replication failed (ruizpardo2022), and the meta-analysis parry2024 finds only a small working-memory effect.

Suggested routines

3,018 adults, a 30, 60 or 90 s hot-to-cold shower for 30 days: self-reported sickness absence was 29 % lower (IRR 0.71), but illness days did not differ. Erratum: PLoS ONE 2018;13(8):e0201978.

Suggested routines

41 adults with type 2 diabetes, two-week crossover periods: walking 10 minutes after each main meal lowered postprandial glucose compared with one 30-minute daily walk (ratio of geometric means 0.88), most clearly after the evening meal (0.78). Tested in a clinical population; transfer to healthy adults is plausible but not shown here.

Suggested routines

Laboratory comparison of reading on a light-emitting eReader versus a printed book in the hours before bed: with the eReader, participants took longer to fall asleep and showed reduced melatonin secretion, later circadian timing and lower next-morning alertness. Tests the light mechanism, not smartphone use as such.

Suggested routines

n = 8, 7-day crossover: protein spread evenly across meals (about 30 g each) gave 25 % higher 24-hour muscle protein synthesis than protein skewed towards dinner. A surrogate measure in a very small sample, with no outcome for muscle mass or strength. Randomisation is not stated in the abstract.

Suggested routines

Muscle protein synthesis −24 % (alcohol with protein) and −37 % (alcohol with carbohydrate) — but at 1.5 g/kg, which the authors themselves put at 12 ± 2 standard drinks. Outside anything that gets logged in practice, so no penalty is applied.

Alcohol and caffeine

400 mg of caffeine six hours before bedtime costs more than an hour of sleep: “Even at 6 hours, caffeine reduced sleep by more than 1 hour” (n = 12, double-blind, placebo-controlled). Carries the mechanism — but not the dose, since a marker knows only a time of day.

Alcohol and caffeine, Suggested routines

Randomised crossover trial, 19 overweight adults aged 45–65: 2-minute walking bouts every 20 minutes lowered the glucose response to a test drink by about a quarter (iAUC 5.2 light, 4.9 moderate versus 6.9 mmol/L·h sitting) and the insulin response by about 23 %. The tested interval is 20 minutes, not an hour.

Suggested routines

CYP1A2 genotype determines the strength of the caffeine response (n = 35 trained cyclists, 6 mg/kg, 40 km time trial). Aellic will never have this genotype; the learned effect captures the phenotype implicitly. Note: an erratum appeared in 2015 (J Int Soc Sports Nutr 12:24, doi 10.1186/s12970-015-0079-6) — two participants had been assigned to the wrong genotype group. The core claim holds, the percentages shift slightly; anyone citing them must use the corrected ones.

Alcohol and caffeine

5,109,185 person-days from 20,968 participants: women are affected more strongly (resting heart rate +2.8 bpm versus +2.4 bpm in men) — but the paper publishes NO multiplier, only absolute sex-separated differences. Hence no sex multiplier in Aellic; the per-person learned effect absorbs it in any case.

Alcohol and caffeine

UK Biobank (n = 60,977, 1,859 deaths): sleep regularity is a STRONGER predictor of mortality than sleep duration — 20–48 % lower all-cause mortality in the upper four SRI quintiles compared with the lowest.

Suggested routines

UK Biobank, median 2.5 h of self-reported daytime outdoor light: each additional hour was associated with greater ease of getting up (95 % CI of the OR 1.46–1.49), less frequent tiredness (0.80–0.82), fewer insomnia symptoms and an earlier chronotype. Self-reported and mostly cross-sectional; the dose unit is hours, and no 30-minute threshold was tested.

Suggested routines

Online survey, n = 793 aged 18–29: eating within 3 h of bedtime was associated with nocturnal awakening (OR 1.61, 95 % CI 1.15–2.27; 1.43, 1.00–2.04 after adjustment) but not with long sleep onset latency or short sleep duration. Cross-sectional and self-reported; the authors call it preliminary.

Suggested routines

Accelerometer data from 4,840 US adults aged 40 or over, mean follow-up 10.1 years: 8,000 steps a day versus 4,000 was associated with lower all-cause mortality (HR 0.49, 95 % CI 0.44–0.55). Step intensity was not associated with mortality once total steps were accounted for.

Suggested routines

36 city dwellers over 8 weeks, repeated measures without a control group: a nature experience lowered salivary cortisol by 21.3 %/hour beyond its normal daily decline. The return per minute was highest between 20 and 30 minutes.

Suggested routines

Within-subject laboratory dose-response study, 55 adults aged 18–30, 10 to 2,000 lux for 5 h in the evening: group ED50 for melatonin suppression was 24.6 lux, with individual values from 6 to 350 lux; 10, 30 and 50 lux delayed melatonin onset by 22, 77 and 109 min. Graded observational only because the registry has no value for a non-randomised experiment.

Suggested routines

Cross-sectional survey, n = 19,806: 120 minutes or more of nature contact a week was associated with good health (OR 1.59) and high well-being (OR 1.23) at 120–179 min; the association peaked at 200–300 min/week, and how the time was split up did not matter.

Suggested routines

Dose-resolved resting heart rate response to alcohol: “HR was increased by 1.4 bpm with low, 4.0 bpm with moderate, and 8.7 bpm with high alcohol intake” (bands: ≤0.25 / >0.25–0.75 / >0.75 g/kg pure alcohol), n = 4,098 employees across 12,411 recording days, selected from an original sample of 111,025 measurement days from 42,086 employees. Preferred over WHOOP's linear extrapolation because it is dose-resolved rather than projected. Firstbeat is the data provider and the employer of two co-authors (Korhonen, Myllymäki); the company's own pages do not carry these bpm values at all.

Alcohol and caffeine, Suggested routines

109 office workers, 7 days of calibrated personal light measurement in winter and summer: high circadian-effective light in the morning was associated with shorter sleep onset latency (especially in winter), stronger circadian entrainment and better sleep quality than low morning light. Observational, with no duration threshold.

Suggested routines

Definition of the Sleep Regularity Index: SRI = 100 × (2 × P(same state at t and t+24 h) − 1), on a scale of −100 to 100. The paper does not ask for “at least 7 days” but for whole MULTIPLES of 7 days — a weekly grid, so that the distribution of weekdays cannot skew the result.

Suggested routines

Health and Retirement Study, n = 3,635: book readers had an adjusted HR of 0.80 for mortality over 12 years compared with non-readers, partly mediated by cognition. Observational and prone to confounding; there is no page-count dose.

Suggested routines

2,315 Finnish men aged 42–60, median follow-up 20.7 years. Compared with one session a week, 2–3 sessions gave an HR for sudden cardiac death of 0.78 (95 % CI 0.57–1.07, not significant); only 4–7 sessions was significant (HR 0.37). Men only, observational.

Suggested routines

52 healthy adults, polysomnography plus a 3-day food diary: nocturnal fat and caloric intake correlated with poorer sleep variables (in women with sleep latency, sleep efficiency and wake after sleep onset). Small, correlational and sex-specific; the authors ask for further studies.

Suggested routines

Retrospective analysis of laboratory data, 116 healthy adults aged 18–30: room light (under 200 lux) versus dim light (under 3 lux) in the 8 h before bedtime delayed melatonin onset in 99 % of individuals and shortened melatonin duration by about 90 min. Graded observational because it is a retrospective comparison across two studies, not a randomised trial.

Suggested routines

Affect words arrange on a circle spanned by two bipolar dimensions, pleasure-displeasure (valence) and arousal-sleepiness. Self-reported mood can therefore be placed as one point on a two-axis field instead of a single good-bad scale. NOT supported by the paper: the 0 to 10 scale onto which Aellic maps both axes for its own statistics, which is a convention.

Journal mood on two axes, valence and energy

Adults aged 25 to under 51: 1,410 ml/day from beverages, 860 ml from solid food and 330 ml oxidation water, 2,600 ml/day in total. This is the beverage figure the 1.5 L routine tracks.

Routine suggestions from your data, Suggested routines

Expert panel using a modified Delphi process on 63 publications from a systematic literature search: consistency of sleep onset and offset timing is important for health, safety and performance. The panel also agreed that, when weekday sleep is insufficient, weekend catch-up sleep may be beneficial, so regularity is not an argument against recovering lost sleep.

Suggested routines

Expert consensus in melanopic EDI at the eye: at least 250 lx during the day (daylight preferred), at most 10 lx in the evening from at least 3 hours before bedtime, and at most 1 lx in the sleep environment. It sets intensities, not a minimum duration of daylight exposure.

Suggested routines

Adults should do 150–300 min of moderate-intensity or 75–150 min of vigorous-intensity aerobic activity per week, plus muscle-strengthening activity at moderate or greater intensity involving all major muscle groups on 2 or more days a week. Replacing sedentary time with activity of any intensity is recommended, but the guideline states the evidence was insufficient to quantify a sedentary threshold or a break frequency.

Routine suggestions from your data, Suggested routines

Age-dependent sleep duration recommendations from an expert panel: 7–9 h for young adults and adults, 7–8 h from age 65. A population default, sound as a cold start, not as an individual target. Full report: Sleep Health 1(4):233–243, doi 10.1016/j.sleh.2015.10.004.

Suggested routines

Consensus by modified RAND Appropriateness Method: adults should sleep 7 or more hours per night on a regular basis to promote optimal health. Sleeping more than 9 hours may be appropriate for some groups; the statement sets a floor, not an optimum.

Routine suggestions from your data, Suggested routines

Strong recommendation to keep free sugars below 10 % of total energy (moderate-quality evidence) and a conditional recommendation to go below 5 % (very low quality). It singles out sugar-sweetened beverages as a source that raises total energy intake and may displace more nutritious foods.

Suggested routines

Adequate Intake of total water (drinking water, beverages of all kinds and food moisture together) of 2.0 L/day for adult women and 2.5 L/day for adult men, valid at moderate temperature and moderate activity (PAL 1.6). Derived from observed intakes combined with desirable urine osmolarity, not from outcome trials. The abstract sets no beverage-only value.

Suggested routines

Publishes only “1 drink → −7 ms HRV” together with a persistence of 74 / 29 / 19 % over three days. The HRV figure is NOT used as a cold start in Aellic — extrapolating it to 5 drinks would be invented linearity.

Alcohol and caffeine

Summarises the revised taxonomy of educational objectives: six ordered cognitive process categories, remember, understand, apply, analyse, evaluate and create, with create at the top, crossed with four knowledge types. It is a framework for writing and classifying learning objectives. NOT supported: that everyday activities can be validly measured on this scale, or that reaching a higher category means better thinking. Aellic uses the order only as a lens for an evening reflection, which is a design decision.

Layers of thinking: Bloom's taxonomy as a lens

37 cohorts, 2,258,029 people: per 1 MET higher fitness a relative risk of 0.89 (95 % CI 0.86 to 0.92) for all-cause mortality; the association did not differ by age or sex. Confirms the slope Aellic takes from Kodama.

Aellic age from sleep, activity and fitness

16 cohort studies: muscle-strengthening activity was associated with 10–17 % lower risk of all-cause mortality, cardiovascular disease, total cancer and diabetes, independent of aerobic activity. The associations were J-shaped, with the largest risk reduction at about 30–60 min a week; the effect of higher volumes is unclear.

Aellic age from sleep, activity and fitness

Participant-level data from 15 cohorts (47,471 adults, 3,013 deaths): compared with the lowest quartile (median 3,553 steps a day), the third quartile (median 7,842 steps) had an adjusted hazard ratio of 0.55. Risk fell progressively up to 6,000–8,000 steps a day in adults aged 60 or over, and up to 8,000–10,000 in younger adults. Observational.

Aellic age from sleep, activity and fitness

46 studies, 1,246,203 people: each 10 beats per minute higher resting heart rate was associated with a relative risk of 1.09 (95 % CI 1.07 to 1.12) for all-cause mortality; risk rose linearly from about 45 bpm. Clinic measurements, substantial heterogeneity and publication bias. Aellic applies the slope to nightly resting heart rate from a wearable, which runs lower than a clinic value.

Aellic age from sleep, activity and fitness

Meta-analysis of 33 studies (102,980 people for all-cause mortality): each 1 MET higher cardiorespiratory fitness was associated with a relative risk of 0.87 (95 % CI 0.84 to 0.90). With 1 MET = 3.5 ml/kg/min this is about -0.04 on the log scale per ml/kg/min, the slope Aellic uses for VO2max.

Aellic age from sleep, activity and fitness

German life tables 2022/2024 (published 22 July 2025), single ages 0 to 100 for men and women. Aellic's own ordinary least squares fit of ln(-ln(1 - qx)) on age over ages 35 to 85 gives a Gompertz slope of 0.0968 per year for men and 0.1004 for women (R² about 0.999, doubling about every 7 years). The fit is poor below about age 30. The slope is Aellic's computation on the official table, not a figure the office publishes.

Aellic age from sleep, activity and fitness

19 people, Apple Watch Series 7 against laboratory VO2max: mean absolute percentage error 15.8 %, root mean square error 8.85 ml/kg/min, ICC 0.47; low fitness was overestimated and high fitness underestimated. At Aellic's slope this error alone is about 3.6 years either way, which is why the Aellic age is always labelled as estimated.

Aellic age from sleep, activity and fitness

UK Biobank (n = 60,977, 1,859 deaths): sleep regularity is a STRONGER predictor of mortality than sleep duration — 20–48 % lower all-cause mortality in the upper four SRI quintiles compared with the lowest.

Aellic age from sleep, activity and fitness

116,221 adults, 30 years of follow-up, self-reported leisure-time activity averaged over up to 15 repeats. Moderate and vigorous activity mutually adjusted and also adjusted for sleep duration, diet, BMI, smoking and alcohol. Moderate minutes a week against 0 to 19: 20 to 74, 0.91; 75 to 149, 0.84; 150 to 224, 0.80; 225 to 299, 0.79; 300 to 374, 0.74; 375 to 449, 0.74; 450 to 599, 0.69; 600 or more, 0.68. Vigorous against none: 1 to 74, 0.87; 75 to 149, 0.81; 150 to 224, 0.79; 225 to 299, 0.77; 300 to 374, 0.78; 375 to 449, 0.73; 450 to 599, 0.76; 600 or more, 0.74. Aellic interpolates between category midpoints and smooths the non-monotone vigorous tail to 0.74 from 375 minutes.

Aellic age from sleep, activity and fitness

Accelerometer data from 4,840 US adults aged 40 or over, mean follow-up 10.1 years: 8,000 steps a day versus 4,000 was associated with lower all-cause mortality (HR 0.49, 95 % CI 0.44–0.55). Step intensity was not associated with mortality once total steps were accounted for.

Aellic age from sleep, activity and fitness

8,171 men: fitness stayed associated after adjusting for physical activity (hazard ratio 0.85 per MET), while leisure-time activity lost its association once fitness was accounted for (0.98, 95 % CI 0.88 to 1.10). Basis for halving the movement group when a device-measured VO2max exists.

Aellic age from sleep, activity and fitness

Nurses' Health Study and Health Professionals Follow-up Study: five low-risk lifestyle factors (never smoking, healthy weight, at least 30 minutes a day of moderate to vigorous activity, moderate alcohol, a high-quality diet) against none gave a hazard ratio of 0.26, about 14 years in rate advancement terms. Aellic's set lacks the two largest factors (smoking and weight), so its total cap stays well below this. An erratum appeared (PMID 30571372).

Aellic age from sleep, activity and fitness

Phenotypic Age equation derived in NHANES III. Its age term divides by 0.09165 per year, which Aellic reads as the slope of the age-only Gompertz model and cites as corroboration of its own slope of about 0.10. Each year of Phenotypic Age had a hazard ratio of 1.09 in both sexes. A correction appeared (PMID 30802240).

Aellic age from sleep, activity and fitness

Definition of the Sleep Regularity Index: SRI = 100 × (2 × P(same state at t and t+24 h) − 1), on a scale of −100 to 100. The paper does not ask for “at least 7 days” but for whole MULTIPLES of 7 days — a weekly grid, so that the distribution of weekdays cannot skew the result.

Aellic age from sleep, activity and fitness

2,798 men: resting heart rate stayed associated with mortality after adjusting for directly measured VO2max and leisure-time activity (16 % per 10 bpm); adding VO2max kept about 67 to 85 % of the log hazard ratio. Basis for weighting resting heart rate at 0.7 when a device-measured VO2max exists.

Aellic age from sleep, activity and fitness

Directly measured VO2peak by decade in 3,816 healthy Norwegian adults. Men: 54.4 (20s), 49.1 (30s), 47.2 (40s), 42.6 (50s), 39.2 (60s), 35.3 (70+); women: 43.0, 40.0, 38.4, 34.4, 31.1, 28.3 ml/kg/min, each with an SD of about 5 to 8. Decline about 3.5 ml/kg/min per decade. Reading the fitness age as the age whose decade mean equals the user's estimated VO2max, interpolated between decade midpoints, is Aellic's construction on top of this table, not a result of the paper.

Aellic age from sleep, activity and fitness

In men the association of leisure-time physical activity with mortality was eliminated after further adjustment for cardiorespiratory fitness; activity and fitness correlated at about 0.49 in men and 0.47 in women. Second basis for halving the movement group next to a device-measured VO2max. Only the abstract was read.

Aellic age from sleep, activity and fitness

EPIC-Norfolk: people with none of four healthy behaviours (not smoking, not inactive, moderate alcohol, plasma vitamin C as a marker of fruit and vegetable intake) had a relative risk of 4.04 against people with all four, which the authors describe as equivalent to being 14 years older. Aellic uses it as an anchor for its total cap and as a check on its age slope (ln 4.04 / 14, about 0.0997 per year).

Aellic age from sleep, activity and fitness

The paper from which the heart rate reserve method emerged: personalise zones through the reserve, that is, take resting heart rate into account rather than maximum heart rate alone. Assessment: the closed formula THR = HRrest + f × (HRmax − HRrest) is a later textbook formalisation, not a formula this small longitudinal study puts forward itself. PubMed carries no abstract; the content is evidenced only through secondary literature.

Aellic age from sleep, activity and fitness

Adults should do 150–300 min of moderate-intensity or 75–150 min of vigorous-intensity aerobic activity per week, plus muscle-strengthening activity at moderate or greater intensity involving all major muscle groups on 2 or more days a week. Replacing sedentary time with activity of any intensity is recommended, but the guideline states the evidence was insufficient to quantify a sedentary threshold or a break frequency.

Aellic age from sleep, activity and fitness

American College of Sports Medicine guidance on exercise intensity: moderate intensity is 40 to 59 % of heart rate reserve, vigorous intensity 60 to 89 %. The table was read as quoted in two open-access reviews (PMC8716584, PMC10248180), not in the paywalled original. Aellic sorts each workout into moderate or vigorous by its average heart rate against these limits.

Aellic age from sleep, activity and fitness

WHOOP's own description of WHOOP Age: nine inputs mapped to published all-cause mortality hazard ratios and converted with 10 × ln(HR), measured against a health-optimised reference so that meeting the recommendations gives the calendar age. Resting heart rate during sleep is neutral at about 60 bpm for men and 64 bpm for women (derivation unpublished). The age uses the last 6 months, the pace of aging the last 30 days projected 6 months ahead on a scale of -1.0x to 3.0x; it updates weekly and unlocks after 21 recoveries in 31 days. Industry document without peer review; its overlap corrections are unpublished. Aellic takes the resting heart rate neutral points, the unlock rule and the pace formula from it.

Aellic age from sleep, activity and fitness

The binding scoring standard for sleep stages. It prescribes that the share of each stage in total sleep time be REPORTED — but it names no normal ranges for those shares.

Total testosterone is best measured in a morning sample, and the harmonised lower limit of the normal range in healthy young men is about 264 ng/dl (9.2 nmol/l).

Defines rate advancement periods: the time by which the rate of a chronic condition is advanced among exposed people, derived from the output of standard regression models. Aellic uses the same idea to express a hazard ratio as years of age. Only the abstract was read.

DOES NOT SUPPORT THE RECOMMENDATION ATTRIBUTED TO IT — and contradicts it. It is not a review but a randomised trial, and its result reads: a deload week in the middle of a 9-week programme “appears to negatively influence measures of lower body muscle strength but has no effect on lower body hypertrophy, power or local muscular endurance”. A recommendation of “every 3–6 weeks” appears nowhere in it. Until 2026-07-25 this source stood in Aellic for the opposite of what it shows; the observed intervals are now carried by rogerson2024.

Suggests against routine 25-hydroxyvitamin D testing in the general adult population, including ages 75 and over, and no longer endorses any target level for sufficiency, so no optimal band is defined for vitamin D here.

Rate advancement periods are often misread as differences in mean survival time or as a shift of the survival curve; they are neither, and they are highly sensitive to how the dependence of rates on age is modelled. Only the abstract was read.

In HFE C282Y homozygotes, transferrin saturation above 45 percent with ferritin above 200 ug/l in women (above 50 percent and 300 ug/l in men and postmenopausal women) is enough to diagnose haemochromatosis; used as the upper end of the transferrin saturation band.

EAU uses a total testosterone below 12 nmol/l (about 350 ng/dl) as the threshold to diagnose late-onset hypogonadism, measured fasting between 07:00 and 10:00 and repeated at least twice; below 8 nmol/l (about 230 ng/dl) is described as clearly reduced.

A serum 25-hydroxyvitamin D of 50 nmol/l (20 ng/ml) is regarded as sufficient for bone health in the general population.

Accredited university laboratory ranges for adults, split by sex and age band where the analyte needs it (for example haemoglobin, haematocrit, ferritin, creatinine, liver enzymes, testosterone). These are the default ranges for most blood markers when a lab sheet prints none.

DOES NOT SUPPORT THE RULE ATTRIBUTED TO IT. The paper measures SD1, SD2, HF and LF — RMSSD does not appear at all — samples weekly over two years in n = 18, and finds the OPPOSITE direction: a RISE in the parasympathetic indices (HF, SD1) measured supine one week earlier went together with a HIGHER risk of illness. A threshold of “≥20 % drop over 3–5 days” appears nowhere in it; a 3–5 day window is not even part of the study design. Until 2026-07-25 this source carried Aellic's deviation rule — that rule is a practitioner heuristic without a clean primary source and now stands as an openly unsourced design decision.

Holotranscobalamin (active B12) is the earliest marker of a negative B12 balance; the review treats values below 35 pmol/l as lowered, and lowered holotranscobalamin with raised methylmalonic acid as a metabolically manifest deficiency.

Reference data for healthy sleepers — and an explicit objection to hard cut-offs: “values often classified as pathological by clinicians, such as a SOL longer than 30 minutes and a SE below 80 %, are well within the bounds of the healthy sample”. Until 2026-07-25 this paper sat under the id “mitterling” by mistake, because its DOI had been entered there.

Method comparison giving the IFCC to NGSP master equation, NGSP % = 0.915 x IFCC % + 2.15. Since IFCC results are reported in mmol/mol, the same relation reads NGSP % = 0.09148 x IFCC mmol/mol + 2.152 (as published by NGSP), the conversion applied when a lab sheet reports HbA1c in mmol/mol.

The primary data behind the “two- to fourfold risk” that gabbett2016 quotes. A negative training-stress balance raised injury risk in the following week (RR 2.2 for internal, 2.1 for external workload), and a balance above 200 % raised it to RR 4.5 (internal) and 3.3 (external) versus a balance of 50–99 %. One sport, one role — elite cricket fast bowlers — which is the ceiling on how far it generalises to Aellic's users.

Practically all persons are sufficient for bone health at a serum 25-hydroxyvitamin D of 50 nmol/l (20 ng/ml) or more; there may be reason for concern above 125 nmol/l (50 ng/ml).

An estimated glomerular filtration rate of 60 ml/min/1.73 m2 or more is the normal-to-mildly-reduced band (G1 90 and above, G2 60 to 89). Chronic kidney disease means abnormal kidney structure or function for at least three months: markers of kidney damage, or a filtration rate below 60; G1 or G2 alone, without damage markers, does not meet the definition.

Serum iron reference ranges of 65 to 175 ug/dl for men and 50 to 170 ug/dl for women.

Lipoprotein(a) below 30 mg/dl (about 75 nmol/l) is the lower-risk range, and above 50 mg/dl (about 125 nmol/l) is the level that adds cardiovascular risk. mg/dl and nmol/l cannot be converted exactly because particle size varies.

“The median time to reach 95 % of asymptote was 66 days, with a range from 18 to 254 days.” A habit takes a median of 66 days to become automatic, with considerable spread — not 21 days. To be cited before any statement of the kind “you have built a habit”. The curve was fitted for 39 of 96 participants.

669 adults: actigraphy measured 6.0 hours of sleep against 6.8 hours reported; people sleeping 5 hours over-reported by 1.2 hours, people sleeping 7 hours by 0.4 hours (correlation 0.47). The classic U-shape around 7 hours comes from self-report, so Aellic takes its sleep duration curve from device-measured data.

14,689 physically active people wearing a commercial wrist device for one year, 4,084,354 person-nights: later exercise timing and higher exercise strain were associated with delayed sleep onset, shorter sleep duration, lower sleep quality, higher nocturnal resting heart rate and lower nocturnal heart rate variability. Regardless of strain, exercise ending 4 hours or more before sleep onset was not associated with changes in sleep. Observational, in an active sample; it supports a distance to sleep onset, not a fixed clock time, and its abstract reports nothing on deep sleep.

European guideline that sets LDL cholesterol goals by cardiovascular risk (below 3.0 mmol/l, 116 mg/dl, may be considered at low risk) and sets no triglyceride goal, but treats triglycerides below 1.7 mmol/l (150 mg/dl) as lower risk.

Proposes an optimal fasting triglyceride level below 100 mg/dl as a marker of metabolic health, not a treatment target; fasting levels below 150 mg/dl are normal, 150 to 199 borderline high, 200 to 499 high and 500 and above very high.

Polysomnography in 100 healthy adults to AASM standard; reports sleep onset latency as age percentile curves and finds no age effect.

Classifies LDL cholesterol below 100 mg/dl as optimal, total cholesterol below 200 mg/dl as desirable and HDL cholesterol below 40 mg/dl as low.

Background only. A nonexercise fitness model built on age, waist circumference, resting heart rate and activity predicts long-term outcomes in the HUNT cohort. Aellic cites it to justify estimating fitness at all; no number from this paper enters the app, and its outcome wording stays out of user copy.

A non-fasting lipid profile is acceptable for routine use. Triglycerides are the value that moves most after a meal (about +0.3 mmol/l, 26 mg/dl), so a non-fasting triglyceride result can read higher than a fasting one. Laboratory reports should flag non-HDL cholesterol from 3.9 mmol/l (150 mg/dl) and non-fasting triglycerides from 2 mmol/l (175 mg/dl); in fasting samples triglycerides are abnormal from 1.7 mmol/l (150 mg/dl).

Age-based and longevity-based translations of a mortality hazard ratio are valid only when the hazard ratio is constant over age and the rates are log-linear in age. Only the abstract was read.

High-sensitivity CRP below 1 mg/l is the low-risk band, 1 to 3 mg/l average and above 3 mg/l high relative cardiovascular risk.

Suggests a serum ferritin threshold of 30 ng/ml or less instead of 15 ng/ml to diagnose iron deficiency in adults without inflammation, menstruation or pregnancy (conditional recommendation, low certainty). With inflammation it recommends measuring ferritin together with transferrin saturation and sets no fixed threshold; ferritin below 100 ng/ml or transferrin saturation below 20 percent may be considered. Read in the ASH visual summary and teaching slides; the journal text was not accessible.

Social jetlag is associated with a higher BMI — and independently of sleep duration.

A survey of practice as it is actually lived: deloads are “integrated into the training programme every 5.6 ± 2.3 weeks”, triggered predominantly when performance stagnates or when soreness and joint complaints increase. Describes what trainees DO — it makes no claim about efficacy.

UK Biobank, 88,282 adults with actigraphy: against 7 hours of device-measured sleep, 5 hours was associated with a hazard ratio of 1.29 (95 % CI 1.09 to 1.52) for all-cause mortality; 8 hours was not associated. Adjusted for accelerometer activity and BMI among others (covariate list read via the PubMed Central page), not for sleep regularity.

“When comparing studies that investigated training muscle groups between 1 to 3 days per week on a volume-equated basis, the current body of evidence indicates that frequencies of training twice a week promote superior hypertrophic outcomes to once a week” (ES 0.49 versus 0.30).

21 trained men, 8 weeks, 3 full-body sessions per week at 3 × 8–12 RM. Three minutes of rest beat one for strength (1RM squat and bench press both significantly higher) and for hypertrophy — with the caveat that the size advantage was site-dependent: significant in the anterior thigh, only a trend at the triceps brachii (p = 0.06). Overturned the old dogma of “short rests for hypertrophy”.

Preliminary upper reference limits at 37 C with the IFCC reference procedures (97.5th percentile, hospital patients aged 17 and over), men and women: ALT 45 and 34 U/l, AST 35 and 31, gamma-GT 55 and 38, CK 171 and 145, LDH 248 and 247 U/l. The app does not use these stricter limits; its liver enzyme defaults follow the DGKL/VDGH consensus (dgkl_vdgh2005).

Fasting plasma glucose below 100 mg/dl and HbA1c below 5.7 percent are the normal range; fasting 100 to 125 mg/dl and HbA1c 5.7 to 6.4 percent mark the intermediate range, and 126 mg/dl or 6.5 percent and above is the diagnostic threshold for diabetes.

The most-cited practitioner reference for HRV metrics and norms; its Table 6 carries the short-term norms from Nunan 2010. Usable only as a cold-start fallback — the spread among healthy adults is too wide to decide anything about an individual.

Effective age: with a log-linear baseline hazard, an exposure with hazard ratio HR corresponds to ln(HR) divided by the per-year log hazard of age, about 10 × ln(HR) when the hazard rises by roughly 1.1 per year from age 30. Valid only under proportional hazards and a hazard that rises exponentially with age. Summing several effects is only safe for mutually adjusted estimates or clearly distinct exposures, and the author warns against claiming that behaviour change takes years off without randomised evidence. Read in full text (open access).

Consumer devices deviate from a multi-channel ECG reference to differing degrees; an absolute RMSSD comparison across device boundaries is therefore invalid.

23 studies comparing a single evening exercise session with no exercise in healthy adults: overall the studies do not support the hypothesis that evening exercise negatively affects sleep, and slow-wave sleep was slightly higher after evening exercise. The exception is vigorous exercise ending one hour or less before bedtime, which might impair sleep onset latency, total sleep time and sleep efficiency. Only hard sessions close to sleep are therefore treated as late, and no claim of less deep sleep rests on this paper.

Interim consensus upper reference limits of the German laboratory societies DGKL and VDGH for the IFCC methods: alanine and aspartate aminotransferase up to 50 U/l in men and up to 35 U/l in women (with pyridoxal phosphate activation), gamma-glutamyltransferase below 60 U/l in men and below 40 U/l in women. The paper itself is paywalled; the values were read in the Roche method sheets that cite it (ALTP and ASTP page 3, GGT-2 page 4). These are the default liver enzyme limits when a lab sheet prints none.

With new reagents introduced on 26 April 2022 the laboratory set gamma-glutamyltransferase upper limits of below 55 U/l for men and below 38 U/l for women (previously below 64 and below 36 U/l); old and new assays agreed very well. Not the source of the app's gamma-GT limit (see dgkl_vdgh2005).

Serum folate below 10 nmol/l (stated as 4 ng/ml) indicates folate deficiency judged by rising homocysteine, and below 6.8 nmol/l (3 ng/ml) judged by macrocytic anaemia; the cut-offs are meant for assessing populations and do not apply in pregnancy.

In apparently healthy adults a serum ferritin below 15 ug/l indicates depleted iron stores, and below 70 ug/l in the presence of infection or inflammation.

Haemoglobin below 12.0 g/dl in non-pregnant women and below 13.0 g/dl in men aged 15 to 65 years defines anaemia; the 2024 guideline kept these adult cut-offs and sets none above 65. Used as a cross-check on the lower ends of the sex-specific haemoglobin ranges.