REF-1371 · Hecksteden et al., 2017 · International Journal of Sports Physiology and Performance · Paper · PubMed

Can a CK blood test tell me whether I have recovered?

Not shown: in 14 elite juniors, CK read against each one's own values sorted Monday (rested) from Friday samples better than a group range. A marker, small study (B-, abstract).

Main caveat: Abstract only. The full text is paywalled, has no Europe PMC copy, and the Ruhr University Bochum repository copy listed by OpenAlex did not respond to three…
limited, mixed evidenceMeasured on highly trained to elite athletes

Hecksteden, Pitsch, Julian, Pfeiffer, Kellmann, Ferrauti and Meyer start from the problem that differences between individuals may limit how accurately group-based reference ranges diagnose muscle recovery, and try to build individual reference ranges with a Bayesian approach comparable to the Athlete Biological Passport.

They chose urea and creatine kinase (CK) as indicators of muscle recovery. For each, prior distributions and repeated-measures SDs came from 883 squad athletes (1,758 data points, 1-8 per athlete, 2013-2015), and the individualisation equations were adapted to discriminate two physiological states, recovered and non-recovered.

The method was tested over 5 consecutive weekly microcycles in 14 elite junior swimmers and triathletes, with blood drawn every Monday (taken as recovered) and Friday, following a repetitive weekly training schedule. At group level, the change in recovery state was confirmed by significant differences in urea and CK and by validated questionnaires.

Group-based reference ranges were derived from the same data set, so as not to overstate the benefit of individualisation. For CK, error rates were significantly lower with individualised classification than with group-based ranges (test-pass error rate P = .008; test-fail error rate P < .001); for urea, numerical improvements did not reach significance.

The authors conclude that individualised reference ranges seem a promising tool for more accurate monitoring of muscle recovery and that a larger panel of indicators should be tested.

Our reading: this is the base's named counter-position on "one CK value cannot show recovery" in PHYS-25 — and it agrees with the "one value" half: it is the athlete's own series, not a single reading against a population range, that carried the information.

It is one small sample of elite juniors with a fixed weekly rhythm, and the outcome is a marker's agreement with that rhythm, not a training decision.

Protocol

Design: method development (priors and repeated-measures SDs from a database) and validation (classification accuracy in a prospective repeated-measures sample).
Prior data: 883 squad athletes, 1,758 data points (1-8 per athlete), 2013-2015 (used to set the priors, not to test the method).
Evaluation sample: n = 14 elite junior swimmers and triathletes; 5 consecutive weekly microcycles; blood sampled every Monday (defined as recovered) and Friday (end of the training week) on a repetitive weekly schedule.
Markers: CK and urea.
Reference standard: the training-schedule state (Monday recovered vs Friday non-recovered), confirmed at group level by significant marker differences and by validated questionnaires (not named in the abstract).
Comparison: individualised (Bayesian) classification vs group-based reference ranges derived from the same data set.
Outcomes: test-pass and test-fail error rates.
Not given in the abstract: the error rates themselves, sensitivity and specificity, the sex split, how many samples each athlete gave, the questionnaires, and funding and competing interests.

Key Results

Item (abstract)Result
Prior database883 squad athletes; 1,758 data points; 1-8 per athlete; 2013-2015
Evaluation sample14 elite junior swimmers and triathletes; 5 weekly microcycles; Monday and Friday samples
Group-level checksignificant Monday-Friday differences in urea and CK and in validated questionnaires
CK, individualised vs group-basederror rates significantly lower: test-pass error rate P = .008; test-fail error rate P < .001
Urea, individualised vs group-basednumerical improvement, not significant
Conclusionindividualised reference ranges "seem to be a promising tool"; a larger panel of indicators warranted

Mechanisms

Measured: CK and urea differed between Monday and Friday at group level; against each athlete's own expected range, CK misclassified the recovery state less often than against a group range.
Authors' interpretation: individual differences in CK limit the accuracy of group reference ranges; an individual range updated from each athlete's repeated values, as in the Athlete Biological Passport, removes much of that between-person spread.
Inference: this is consistent with PHYS-25's premise that CK varies widely between people — the reason a single value against a group range misleads — and adds that within-athlete series can carry a signal; it does not show that the signal is about function, since the reference standard was the day of the week, not force, performance or injury.

Practical Application

Can a CK blood test tell me whether I have recovered? Not shown: in 14 elite juniors, CK read against each one's own values sorted Monday (rested) from Friday samples better than a group range. A marker, small study (B-, abstract).
What it is good for as a source: the counter-position to "CK cannot show recovery", in its precise form — against the athlete's own baseline series, CK classification improved; against group ranges it was worse.
What it does not show: that CK predicts performance, injury or readiness for a session, or that one value without a personal baseline is informative; error-rate sizes are not given in the abstract.
How a principle should cite it: "against the athlete's own repeated values CK sorted recovered from non-recovered better than group ranges (Hecksteden 2017, 14 elite juniors, from the abstract)".
What it means for training: Not from this paper; the base's own practice: judge readiness by function, feel and how a known session goes (SYNTH-70, SYNTH-30); a blood marker is read, if at all, as a within-athlete trend, never against a population range (an extension of SYNTH-91's same-athlete trend rule, which is written for device readings).

Limitations

Abstract only. The full text is paywalled, has no Europe PMC copy, and the Ruhr University Bochum repository copy listed by OpenAlex did not respond to three attempts on 2026-10-28. Error rates, the classification threshold, the questionnaires, the sex split and the funding and competing interests could not be read.
Small evaluation sample (14 athletes) of one type (elite junior swimmers and triathletes) with a fixed weekly rhythm; the Monday "recovered" label is an assumption of the schedule, checked only at group level.
A marker outcome. Agreement with a training-day label is not the athlete's outcome; no performance, force or injury outcome.
Group ranges from the evaluation sample itself. The authors built them this way to avoid overstating the benefit of individualisation, which makes the group comparator a best case for group ranges; the individual method's priors came from the separate 883-athlete database.
How this entry was checked and changed
Why this level

a test validation study (B by the rubric) with a serious limitation beyond its design: the evaluation sample is 14 athletes, the outcome is a blood marker's classification, not performance, and the error rates themselves are not given in the abstract (only P values). The full text was not read: no Europe PMC copy, and the OpenAlex-listed university repository copy did not respond