REF-1258 · Goods et al., 2023 · European Journal of Sport Science · Paper · PubMed

Can CORE tell me my core temperature in the heat?

Not to a threshold: in 19 elite hockey players, CORE read 0.06-0.34 °C below an ingested pill on average, 41-60% of readings were >0.3 °C off, and the gap widened as core rose.

B-limited, mixed evidenceMeasured on elite athletes

Goods et al. (Murdoch University, Western Australian Institute of Sport, the Australian Institute of Sport, Hockey Australia and the University of Western Australia) asked whether the CORE wearable sensor, which estimates core temperature from the skin, agrees with an ingested telemetric pill when core temperature is high.

Nineteen players of the Australian national women's field hockey squad wore CORE and took a BodyCAP pill through 4 training sessions in hot conditions at a heat camp before the 2020 Olympic Games. Lin's concordance was "poor" in every session.

CORE read lower than the pill in all sessions, by 0.06-0.34 °C on average, with what the abstract calls wide mean 95% confidence intervals of ±0.35 to ±0.56 °C. A smoothed regression showed the error was not constant: CORE under-read more as core temperature rose. In every session 41-60% of all samples differed by more than ±0.3 °C.

The authors conclude that CORE is not a valid alternative to a telemetric pill, particularly during exercise in the heat when core temperature is expected to be high.

Our reading: in elite athletes training in the heat, a CORE reading was within ±0.3 °C of the pill about half the time or less, and it erred low exactly where a heat-training target or a safety limit sits; from the abstract, so the sensor version and the conditions are not known.

Protocol

Who: n = 19 female field hockey players in the Australian national squad.
Setting: a team-sport heat training camp before the 2020 Olympic Games; 4 sessions in hot conditions (the abstract does not say whether outdoors or in a chamber, or give temperature and humidity).
Devices: CORE wearable temperature sensor (index) and BodyCAP telemetric temperature pill (reference), both recording during each session.
Analysis: Lin's concordance correlation coefficient; mean bias with 95% intervals; locally estimated scatterplot smoothing (LOESS) of the error against core temperature; the share of samples beyond ±0.3 °C.
Not in the abstract: CORE hardware generation and firmware; where and how CORE was worn; pairing with a heart-rate strap; pill ingestion timing (which affects pill readings); sampling rate and number of samples; the core-temperature range reached; the concordance values; whether the ±0.35-0.56 °C intervals are confidence intervals of the bias or limits of agreement.
Funding and interests: not given in the abstract or the PubMed and Crossref records (full text not read).

Key Results

OutcomeResult (abstract)
Concordance (Lin's CCC)"Poor" agreement in all 4 sessions
Mean bias, CORE minus pill−0.06 °C to −0.34 °C (CORE lower in every session)
Width around the bias"Mean 95% confidence intervals" (the abstract's term) ±0.35 °C to ±0.56 °C
Error vs core temperatureNon-linear; greater under-reading as core temperature rose (LOESS)
Samples beyond ±0.3 °C41-60% of all samples in each session
Reading the intervals: the abstract calls ±0.35-0.56 °C "mean 95% confidence intervals". If they were computed over all data samples (the abstract gives no count), intervals of that size around a mean bias would be unusually wide, and with 41-60% of samples beyond ±0.3 °C they behave like limits of agreement for single readings (our reading; the full text, not read, would settle it).
The ±0.3 °C threshold is the authors' criterion for an acceptable error; it was exceeded by 41-60% of samples.

Mechanisms

Measured: the gap between CORE and the pill was not a fixed offset: it grew, with CORE lower, as core temperature rose.
Authors' interpretation: wearable sensors could overcome the practical limits of pills and probes, but their accuracy at elevated core temperatures has been questioned, and in this study it did not hold during exercise in the heat.
Inference: CORE estimates core temperature from skin temperature and heat flux through the skin with an algorithm; sweating, airflow, sun and the changing skin-to-core gradient during hard exercise in the heat all act at the skin, so an estimate built there can lag or flatten exactly when the core is rising fastest. Not tested in the abstract.
Inference: if under-reading grows as core temperature rises, the CORE curve also rises less steeply than the true one near the top, so even a within-session trend understates the late rise.

Practical Application

Can CORE tell me my core temperature in the heat? Not to a threshold: in 19 elite hockey players, CORE read 0.06-0.34 °C below an ingested pill on average, 41-60% of readings were >0.3 °C off, and the gap widened as core rose.
Do not gate a heat session or a safety stop on a CORE number (B-, from the abstract): a reading of 38.5 °C could be several tenths of a degree from the truth, and in this study the error ran low at high temperatures — the true core may be hotter than shown. Not from this paper; the base's own practice: a heat session ends on how the athlete feels and functions (confusion, staggering, nausea, unusual weakness), not on a sensor number.
Heat-training targets: SYNTH-04's overdressing route aims to "hold core ≥38.5 °C", a target it labels as practice taken from a rectal-probe protocol (REF-3489), and its own verification notes that a home athlete has no way to measure it.CORE does not close that gap: it cannot place an athlete within ±0.3 °C of 38.5 °C reliably, and by this study's direction, holding CORE at 38.5 °C likely means a true core at or above that — more heat than shown, which is a safety concern rather than a training bonus (inference).
What CORE can still do (inference, not tested here): show that a session is getting hotter, and compare similar sessions on the same sensor in the same conditions; pair it with heart rate, perceived exertion and thermal comfort.
Be exact about what was tested: one CORE version in use before the Tokyo Games (2020-21) on elite female hockey players training in hot conditions, against a pill; indoor trainer riding, later hardware (CORE 2) and firmware were not tested.

Limitations

From the abstract only. The full text (Wiley, open access under CC BY-NC-ND) could not be retrieved by the permitted route (a bot check blocks automated download; the UWA repository record holds no file); the sensor version, placement, conditions, pill timing, sample numbers, concordance values, funding and conflicts are unknown.
One device version. The CORE sensor in use before the Tokyo Games (2020-21); the maker has since released new hardware and firmware.
Who and what. Elite women in a team sport with intermittent running in the heat; not endurance cyclists or runners in steady exercise, not indoor riding.
Criterion. An ingestible pill is a valid field reference but reads the temperature where it sits in the gut and is affected by recent drinking and ingestion timing; how this was handled is not in the abstract.
Interval wording. The ±0.35-0.56 °C values are called confidence intervals in the abstract; whether they are limits of agreement is not stated.
How this entry was checked and changed
Why this level

from the abstract; a validation study with n = 19 elite athletes and a valid criterion, which the rubric places at B; held at B- because only the abstract could be read (the open-access full text is behind a bot check), so the sensor version, firmware, placement, conditions, sample counts and the funding and conflict statements are unknown