Journal Logo


Improving the Diagnosis of Nonfunctional Overreaching and Overtraining Syndrome


Author Information
Medicine & Science in Sports & Exercise: December 2019 - Volume 51 - Issue 12 - p 2524-2530
doi: 10.1249/MSS.0000000000002084
  • Free


Nonfunctional overreaching (NFO) and the “overtraining syndrome” (OTS) are both characterized by performance decline, high levels of fatigue and psychological and hormonal disturbances. It is believed that the symptoms of OTS are more severe than those of NFO. However, the distinction between NFO and OTS is difficult to establish, and must be made based on exclusion diagnosis (1). The “2006 consensus statement” of the European College of Sport Science (ECSS) and “2012 joint consensus statement” of ECSS and American College of Sports Medicine (ACSM) indicated that time needed to recover is the most discriminating factor between NFO and OTS (1,2). Fully recovering from NFO needs weeks to months, whereas recovering from OTS takes months to years, and the athletes can probably never reach their best level again (1). Therefore, there is a strong need for relevant tools for the diagnosis of NFO versus OTS (3).

Most literature agree that NFO and OTS must be seen as a continuum involving disturbance, adaptation and finally maladaptation of the hypothalamic–pituitary–adrenal (HPA) axis (1). In a recent systematic review, it was concluded that basal hormonal levels of the HPA axis cannot be used as predictors of NFO and OTS (4). The gold standard functional test to evaluate the whole HPA axis integrity in response to a stressful situation is the insulin tolerance test (ITT), where hypoglycemia is used as the trigger for stress (4). Two studies (5,6) found blunted adrenocorticotrophic hormone (ACTH) and cortisol levels in athletes with NFO and OTS. Although this test is recommended by endocrine societies, it is not exercise specific. Studies that evaluated the hormonal response to a maximal exercise test in NFO/OTS showed a blunted hormonal response to acute exercise in OTS and NFO versus normal athletes. However, the difference between OTS and NFO could not be made and therefore the Training Optimization (TOP) test seems to be more appropriate to distinguish NFO and OTS (4). The TOP test, which is a two-bout maximal incremental exercise test protocol, was created to detect changes in the typical HPA axis-related hormones (cortisol, ACTH, prolactin [PRL], and human growth hormone [GH]) in response to two exercise tests separated by 4 h (7). In a small sample size (n = 10), it was shown that the TOP test may be a useful tool for diagnosing NFO and OTS, showing an overshoot of ACTH and PRL reaction after the second exercise bout in NFO, and a suppression in OTS (3,7). A cutoff value of 200% for the ACTH and PRL response to the second exercise test was shown to be discriminating between NFO and OTS with high sensitivity (>80%). Cortisol and GH responses to the second exercise bout were much less sensitive, as were resting values of these HPA axis hormones (3). However, more data were needed before this test can be used as a gold standard, and an update of the study results was needed. In the last 14 yr, 100 patients underwent a TOP test in our laboratory, and the diagnosis of NFO and OTS was made based on (i) exclusion criteria (1), (ii) personal history of underperformance, (iii) visual inspection of the differences in hormonal responses of the two tests during the TOP test, and (iv) psychological disturbances observed in the profile of Mood State (POMS) questionnaire (8–10).

It was the aim of this article to reevaluate the cutoff values for hormonal responses, proposed by Meeusen et al. (3) on our existing database. We aimed to simplify and optimize the distinction between OTS and NFO by developing a multivariate approach (discriminant analysis [DA]) including hormonal and psychological changes during the TOP test.



Between 2004 and 2018, 100 subjects with or without (healthy controls) complaints of underperformance and fatigue underwent a TOP test. A physical examination, using the check list proposed by Meeusen et al. (2), to exclude other possible causes of the decrease in performance, preceded the TOP test. The detailed protocol of the TOP test is described by Meeusen et al. (3). All subjects signed a written informed consent that results of the TOP test could be used for data analysis anonymously. The study protocol was approved by the research ethical committee of the Vrije Universiteit Brussel (B.U.N. 143201837073) and the study was conducted according to the Declaration of Helsinki (1964).

TOP Test Protocol

The TOP test protocol was performed as described in the original article on OTS and NFO diagnosis (3). The athletes underwent two incremental graded exercise tests until exhaustion, with a 4 h rest in between. One hour before each test, the athletes received a standardized meal (2315 kJ, 73% carbohydrate, 19% protein, 8% fat). Athletes arrived in the laboratory at 7:00 am after an overnight fast. The first blood sample was collected as they arrived. Before and after both exercise tests, blood samples for cortisol, ACTH, PRL, and GH were drawn. The exercise tests were performed on a cycle ergometer (Lode Excalibur Sport, Groningen, The Netherlands), on their own bike frame mounted on a Cyclus 2 Record trainer (RBM elektronik-automation GmbH, Leipzig, Germany) or on a treadmill (Ergo ELG 55; Woodway, Weil am Rhein, Germany) depending on the sport. Subjects wore a heart rate monitor (Polar Accurex Plus, Kempele, Finland) for determination of HRmax throughout the exercise tests. After each exercise test, 20 μL of blood was drawn from the right earlobe to determine maximal blood lactate concentration ([La]max) with enzymatic analysis (EKF; Biosen 5030, Barleben, Germany). The blood samples were collected in prefrozen 4.5-mL K3 EDTA vacutainer tubes (Becton Dickinson Vacutainer System Europe, Plymouth, UK) and immediately centrifuged at 3000 rpm (Minifuge 2, Heraeus, Germany) for 10 min, and plasma was frozen at −20°C until further analysis. Samples were assayed via RIA for cortisol (DiaSorin, Stillwater, MN), ACTH (Nichols Institute Diagnostics, San Juan Capistrano, CA), PRL (Roche Diagnostics, Mannheim, Germany) and GH (Pharmacia & Upjohn Diagnostics, Uppsala, Sweden). Hormonal concentrations were expressed both in absolute (average (SE)) and relative values. For the relative hormone concentrations, both pretest values were set at 100%; both posttest values were calculated by dividing through the pretest value and multiplying by 100% (3).

Data Analysis

Cutoff values

Hormonal changes were expressed as relative percentages, where pretest values were set at 100% and posttest values were calculated as ((posttest/pretest) × 100). Sensitivity of cutoff values to detect NFO or OTS, proposed by Meeusen et al. (3), was calculated for the existing database (n = 39 for NFO and n = 15 for OTS). Sensitivity was calculated as number of correct OTS or NFO diagnoses based on the cutoff values, divided by the number of OTS or NFO diagnosed by the two experts. Hormonal responses to the second exercise test were compared between normally trained athletes, NFO and OTS by one-way ANOVA. Tukey post hoc testing, by using the Bonferroni correction, was performed to compare normally trained athletes and NFO and NFO and OTS.

Discriminant analysis

In DA, the canonical coefficient of each independent variable is derived in such way that the group means on the functions are as different as possible. A discriminant function was calculated to distinguish between two diagnoses (NFO vs OTS), and acts as projections of the data onto a dimension that best discriminates between the groups. Significant discriminant functions were interpreted by examining Wilks’ λ, eigenvalues, and standardized canonical discriminant function coefficients (SDC). Wilks’ λ measures the deviation within each group with respect to the total deviation. Eigenvalues describe how much discriminating ability a function possesses, whereas the percentage of variance describes the proportion of discriminating ability of the variables found in a given function. The SDC indicate the importance of each variable contributing most to the function. Positive predictive values (PPV) and negative predictive values (NPV) for OTS were calculated as (true positive prediction (OTS)/(true + false positives)) (PPV) and (true negative prediction (NFO)/(true + false negatives)) (NPV).

A canonical linear DA, entering all independent variables together, was used to determine whether NFO and OTS could be distinguished by the variables measured in the TOP test. Discriminant analysis 1 (DA1) aimed to distinguish between NFO and OTS, excluding the subjects without complaints of underperformance and fatigue. Discriminant analysis 2 (DA2) used a forward stepwise procedure to determine the combination of variables yielding the greatest discrimination. The individual variable providing the greatest univariate discrimination was selected first and then paired with each of the remaining variables one at a time. Discriminant analysis 3 (DA3) also included the variables “sex” and “age,” because the HPA axis is known to vary on aging and sex differences (11).



One hundred well-trained male (n = 74) and female (n = 26) athletes performed a TOP test (mean ± SD; age: 27 ± 8 yr; body mass index: 21.6 ± 1.9). Subjects were active in different sports, namely cycling (n = 40), running (n = 8), triathlon (n = 6), skating (n = 8), gymnastics (n = 2), soccer (n = 2), volleyball (n = 2), swimming (n = 2), rowing (n = 1), and climbing (n = 1). Twenty-nine subjects were military soldiers performing several sports.

On the basis of the diagnosis of the experts, 42 athletes were classified as normally trained, 39 athletes were diagnosed with NFO and 15 athletes with OTS (four athletes were “recovering” from NFO in a follow-up test and were excluded from further analysis). Data of the normally trained athletes was only used in the one-way ANOVA to compare the hormonal changes in the three groups. These athletes were recruited as “normal” for experimental research and performed a TOP test, but not on their own request as a consequence of unexplained underperformance. In the DA, data of normally trained athletes was not included as the DA aimed to make prediction between NFO and OTS for people who suffer from unexplained underperformance.

Cutoff values

Cutoff values to discriminate NFO and OTS, as determined by Meeusen et al. (3) based on a database of n = 10, were as follows: 200% increase to second exercise test for PRL, ACTH, and cortisol, 1000% increase to second exercise test for GH. Table 1 shows the sensitivity for detecting NFO and OTS on a database of n = 54.

Sensitivity for detecting NFO vs OTS based on the cutoff values determined by Meeusen et al. (3).

One way ANOVA showed a significant difference in PRL response to the second exercise test between normally trained athletes, athletes diagnoses with NFO and OTS (Table 2). Post hoc t-tests showed a significant lower PRL response in OTS compared with NFO. No significant differences in cortisol, ACTH, and GH response to the second exercise test were found (Fig. 1).

Hormonal responses in second exercise test in normally trained, NFO and OTS athletes expressed as relative changes (%) compared to status before the second exercise test.
Relative changes of cortisol, ACTH, PRL and GH responses to the TOP test for normally trained (solid lines), NFO (dashed lines) and OTS (pointed lines) group. Data are presented as percentage increase from both baseline values (SE) of the mean. *Significant main effect of diagnosis P < 0.05.

Discriminant analysis

DA1 aimed to distinguish between NFO and OTS (n = 52). The overall χ2 test was significant (Wilks’ λ = 0.32, χ2 = 44.16, df = 22, P = 0.003), and the eigenvalue of the function was 2.10. The ACTH response to the second exercise test (SDC = 1.00) made the greatest contribution to the score of the function (− score for OTS vs + score for NFO), followed by POMS index of fatigue after the first exercise test (SDC = 0.97) and the ACTH responses to the first (SDC = −0.75) and PRL response to the first (SDC = −0.70) and second exercise test (SDC = 0.68) (Table 3). DA1 correctly predicted 98.1% of the cases, as NFO or OTS (NPV = 100%, PPV = 93%) (Fig. 2). This number decreased to 76.9% after cross-validation (leave-one-out classification) (NPV = 93%, PPV = 64%).

Standardized canonical discriminant function coefficients of DA1 discriminating between NFO and OTS including all variables measured during the TOP test.
DA1 scatter plot for discriminating between NFO and OTS using all variables measured in the TOP test.

DA2 used a stepwise method, only including the most important variables discriminating between NFO and OTS. Overall χ2 test was significant (Wilks’ λ = 0.52, χ2 = 31.07, df = 4, P < 0.001), and the eigenvalue of the function was 0.91. The significant function included four variables, being the PRL response to the first (SCD = 0.84) and second (SCD = −1.09) exercise test, GH recovery value (SCD = 0.42), and the POMS feeling of tension after the second exercise test (SCD = 0.693). A more + value of the function indicates a state of OTS, whereas a more − value of the function suggests a state of NFO (Fig. 3). DA2 correctly classified 84.9% of the cases as NFO or OTS (NPV = 94%, PPV = 67%). After cross-validation, classification score reached 83% (NPV = 94%, PPV = 63%).

DA2 scatter plot for discriminating between NFO and OTS using a stepwise method, thereby including only the most discriminating variables.

Similar to DA1, DA3 aimed to discriminate between NFO and OTS based on relative changes in hormone responses and psychological changes to the two exercise tests, but also added sex and age as variables (n = 52). The overall χ2 test was significant (Wilks’ λ = 0.29, χ2 = 47.27, df = 24, P = 0.003), and the eigenvalue of the function was 2.47. The ACTH response to the first (SDC = 0.87) and second exercise test (SDC = −0.81) and PRL response to the first (0.82) and second (SDC = −0.86), together with POMS index of fatigue after the first exercise test (SDC = −1.12) were the most discriminating factors. DA3 correctly predicted 98.1% of the cases, as NFO or OTS (NPV = 100%, PPV = 93%). This number decreased to 76.9% after cross-validation (NPV = 91%, PPV = 55%).

The following equation can be used to determine the likelihood of an athlete to be classified as NFO or OTS.

Group centroids are located at 0.863 for NFO and −2.34 for OTS. When function 1 < −1.48, athletes seem more likely to be classified as OTS, whereas >1.48 suggests a larger chance for NFO, which can support the differential diagnosis of NFO versus OTS.


The present study confirmed previous findings that NFO and OTS are characterized by hormonal and psychological disturbances, measured during the TOP test. The ACTH and PRL responses to the second exercise test in the TOP test are most indicative for the diagnosis of NFO and OTS. However, the sensitivity of previously defined cutoff values (>200% for NFO and <200% for OTS) is too low to use these values as biomarkers for NFO and OTS, without taking into account other psychological and physiological factors. A DA including hormonal and psychological variables measured during the TOP test, can successfully discriminate between NFO and OTS and can be used as a supportive tool in making future diagnoses. The most discriminating variables in this analysis are (in descending priority) PRL response to the second and first test, ACTH response to the second and first tests, with additionally POMS feeling of fatigue after the first test.

Cutoff values

Recently, Cadegiani and Kater (2017) pleaded for the establishment of reliable cutoff values for hormonal responses, as biomarkers for NFO and OTS. Cutoff values to detect OTS (vs healthy athletes) were proposed for the cortisol and ACTH response during and after ITT. (6) Although these values had relatively high predictive value, they could not discriminate between NFO and OTS. Cutoff values for the hormonal responses to the second exercise test during the TOP test were defined by Meeusen et al. (3), showing the highest sensitivity for ACTH and PRL in discriminating NFO and OTS. Even though the sensitivity of these cutoffs decreased when applied to the database of 2018 (n = 100), the ACTH and PRL responses to the second test were still the most sensitive, when compared to the cortisol and GH response. These results are in line with the results of the recent systematic review of Cadegiani (2017) (4), showing that the acute hormonal responses in ACTH, GH, and PRL to a single-exercise test or to the TOP test tend to be altered in NFO and OTS, whereas cortisol and plasma catecholamines show conflicting results.

Discriminant analysis

A multivariate approach (DA1) using hormonal and psychological responses induced by the TOP test successfully discriminated between NFO and OTS (98.1%). The ACTH and PRL responses were the most discriminating factors: an exaggerated response to the second test increased the chances for a diagnosis of NFO, whereas larger hormonal responses to the first test and blunted responses to the second test were indicative for OTS. Taking into account age and sex of the subjects, the DA showed identical results, indicating that no sex-specific or age-specific DA is needed to correctly predict the diagnosis. These results confirm previous studies showing that NFO is characterized by a hormonal overshoot in PRL and ACTH, and OTS by a repression of the hormonal response and relative failure of the HPA axis (3,4).

The ACTH and PRL responses to both the first and second tests were important variables to discriminate between NFO and OTS. Therefore, a TOP test protocol is preferable above a single maximal test to diagnose the stage of the spectrum of underperformance and HPA axis dysfunction. The HPA axis consists of stimulating forward and inhibitory feedback loops involving the brain, the hypothalamus, pituitary, and adrenal glands which produce glucocorticoids. This study showed that cortisol responses were not altered in NFO and OTS. Blunted ACTH and PRL responses to the second test, combined with a normal cortisol response, provide further proofs that the neuroendocrine dysfunction associated with NFO and OTS, is located in the hypothalamus and pituitary and not at the adrenals (3). A recent study also confirmed these results by using not-exercise specific tests. Athletes with OTS showed a normal response on the cosyntropin stimulation test (CST), a test directly stimulating the adrenal glands, when compared to healthy athletes. However, an impaired ACTH response to the ITT in OTS was seen in OTS compared to healthy athletes, also indicating the HPA “abnormalities” are located centrally (hypothalamus or pituitary) (6). This study also showed improvements in HPA axis sensitivity of healthy athletes compared to healthy untrained subjects, suggesting that in OTS, these positive adaptations of the HPA axis to exercise are lost, which may explain the decreased performance.

For the first time, we showed that the “recovery rate of resting GH (pre2/pre1)” is different between NFO and OTS. A higher pre2/pre1 value, meaning that the exercise-induced increase in GH remained elevated after 4 h of rest, indicated a higher chance for OTS. High-intensity exercise is a powerful stimulus for GH secretion by the pituitary gland, but GH normally returns to baseline levels already 30 min upon cessation of exercise (12). The elevated GH level at the start of the second exercise test also likely represents a hypersensitive and thus prolonged and exaggerated reaction of the HPA axis in response to the first exercise test.

Besides the hormonal disturbances, some psychological changes (reported as POMS negative mood changes) were important factors in the DA to discriminate between NFO and OTS. Elevations in negative mood of fatigue after the first test and tension after the second test were different in NFO and OTS, providing further support for the efficacy of psychological assessments in the TOP test.


The difficulty to recruit athletes affected with OTS or NFO leads to small sample sizes and thus low statistical power of most studies examining OTS and NFO. This study is the first presenting results with a larger sample size (n = 100), but the number of athletes with OTS (n = 15) was still relatively low. As a consequence, after a leave-one-out cross-validation, NPV was still relatively high (>90%), but the PPV for OTS decreased to a large extent (to <70%). A higher number of OTS athletes might result in an increase of the PPV. All athletes diagnosed with NFO or OTS in this study (n = 54), suffered from unexplainable underperformance and underwent a TOP test on their own request. Thus, they were not part of a study where a state of overreaching was induced by implementing a vigorous training protocol. We could not “prove” the decline in performance, still seen as the golden standard for the diagnosis of NFO and OTS, as we did not perform baseline (pre-NFO or pre-OTS) measurements, and thus, we were forced to use other evidence for the declined performance, such as competition results, results of field tests, and power data.

We acknowledge that this DA to predict the diagnosis of OTS and NFO was based on the diagnoses made by the experts in the field, who based the diagnoses not only on hormonal and psychological responses during the TOP test, but also on training and performance history, duration of underperformance and personal factors that cannot be quantified. This implements that the proposed DA can be used as a tool to optimize the diagnosis, but needs to be used in combination with a medical and personal examination. Moreover, we must keep in mind that this DA can predict the diagnosis of (future) cases, but that this DA itself is based on previously made diagnoses.


To further validate the TOP test, future research should try to link acute hormonal responses to the TOP test with functional tests standardized by endocrine societies, such as the ITT. On the basis of the findings of this current study, we acknowledge that additional data are necessary to further substantiate the TOP test as the golden standard for NFO and OTS (differential) diagnosis. In addition, it would be interesting and valuable to adopt other valid and sport-specific questionnaires on mood, stress, recovery, and physical symptoms and to compare these results with changes in the hormonal responses and the POMS.

Considering female athlete specificities, it should be taken in account that the female menstrual cycle and the use of hormone contraceptives could play a role. The effect of oral contraceptives on the HPA axis is equivocal, depending on the dosage and the composition of the contraceptive. A study comparing the effect of the TOP test in nonoral contraceptive users versus different types of oral contraceptive users on basal and preexercise/postexercise values of cortisol, ACTH, GH, PRL in different moments of the menstrual cycle should be undertaken to clarify specific female athlete-related issues (13,14).


This study provides further evidence that ACTH and PRL responses during the TOP test are sensitive markers to discriminate between NFO and OTS. We propose a DA, including hormonal and psychological responses during the TOP test, as a useful tool to optimize the diagnosis of NFO and OTS.

The results of the present study do not constitute endorsement by ACSM.

Contribution Statement: L. B. and R. M. contributed to planning experimental design, data gathering, data analysis and manuscript revision. L. D. and R. V. contributed to data analysis and manuscript writing. N. T. contributed to data gathering and data analysis. K. B. contributed to planning experimental design and data analysis.

Data Sharing Statement: Original data can be shared by e-mail upon request to the last author.

All authors confirm that they do not have any financial and personal relationships with other people or organizations that could inappropriately influence (bias) this work. No conflicts of interest to declare. The results are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. No source of funding to declare.


1. Meeusen R, Duclos M, Foster C, et al. Prevention, diagnosis, and treatment of the overtraining syndrome: joint consensus statement of the European College of Sport Science and the American College of Sports Medicine. Med Sci Sports Exerc. 2013;45(1):186–205.
2. Meeusen R, Duclos M, Gleeson M, Rietjens G, Steinacker J, Urhausen A. Prevention, diagnosis and treatment of the Overtraining Syndrome—ECSS Position Statement ‘Task Force.’ Eur J Sport Sci. 2006;6(1):1–14.
3. Meeusen R, Nederhof E, Buyse L, Roelands B, de Schutter G, Piacentini MF. Diagnosing overtraining in athletes using the two-bout exercise protocol. Br J Sports Med. 2010;44(9):642–8.
4. Cadegiani FA, Kater CE. Hormonal aspects of overtraining syndrome: a systematic review. BMC Sports Sci Med Rehabil. 2017;2(9):14–29.
5. Barron JL, Noakes TD, Levy W, Smith C, Millar RP. Hypothalamic dysfunction in overtrained athletes. J Clin Endocrinol Metab. 1985;60(4):803–6.
6. Cadegiani FA, Kater CE. Hypothalamic-pituitary-adrenal (HPA) axis functioning in overtraining syndrome: findings from endocrine and metabolic responses on overtraining syndrome (EROS)-EROS-HPA Axis. Sports Med Open. 2017;3(1):45–56.
7. Meeusen R, Piacentini MF, Busschaert B, Buyse L, De Schutter G, Stray-Gundersen J. Hormonal responses in athletes: the use of a two bout exercise protocol to detect subtle differences in (over)training status. Eur J Appl Physiol. 2004;91(2–3):140–6.
8. Nederhof E, Zwerver J, Brink M, Meeusen R, Lemmink K. Different diagnostic tools in nonfunctional overreaching. Int J Sports Med. 2008;29(7):590–7.
9. Morgan WP, Costill DL, Flynn MG, Raglin JS, O’Connor PJ. Mood disturbance following increased training in swimmers. Med Sci Sports Exerc. 1988;20(4):408–14.
10. Morgan WP, Brown DR, Raglin JS, O’Connor PJ, Ellickson KA. Psychological monitoring of overtraining and staleness. Br J Sports Med. 1987;21(3):107–14.
11. Lim CT, Khoo B. In: Feingold KR, Anawalt B, Boyce A, et al, editors. Normal Physiology of ACTH and GH Release in the Hypothalamus and Anterior Pituitary in Man. In Endotext; 2000.
12. Gough L, Castell LM, Gatti R, Godfrey RJ. Growth hormone concentrations in different body fluids before and after moderate exercise. Sports Med Open. 2015;2:30–6.
13. Burke CW. The effect of oral contraceptives on cortisol metabolism. J Clin Path. 1969;23(3):11–8.
14. Hertel J, König J, Homuth G, et al. Evidence for stress-like alterations in the HPA-axis in women taking oral contraceptives. Sci Rep. 2017;7(11):14111.


Copyright © 2019 by the American College of Sports Medicine