Blood Test Score Outperforms GPS Data in Predicting Football Injuries Within Ten Days

Professional football clubs spend millions on GPS vests, accelerometer data, and workload dashboards, yet players keep tearing hamstrings and rolling ankles in training sessions that looked, on paper, perfectly calibrated. A new study from a team of German sports scientists and intensive care physicians suggests the missing ingredient may not be on the pitch at all, but circulating in the players’ blood. By combining measurements of inflammatory cytokines with markers of mitochondrial vitality, the researchers built a simple six-point injury score that dramatically outperformed conventional workload metrics at flagging which players would be injured within the following ten days.

The study, published in Frontiers in Sports and Active Living, followed twenty-one professional male footballers from a German 2. Bundesliga club across the 2023/2024 pre-season and early competitive phase. Between early July and early August 2023, each player provided venous blood samples immediately before and directly after selected training sessions, while wearable Catapult tracking devices recorded their external training load. Over the five assessed ten-day follow-up windows, during which the team played six matches, six of the twenty-one players sustained a clinically relevant non-contact or indirect-contact injury requiring evaluation by the team physician. The injury spectrum included muscle injuries graded under the British Athletics Muscle Injury Classification as well as ligamentous and joint pathologies.

The researchers first asked whether the club’s existing workload data could have predicted those injuries. They normalized total player load, an accelerometer-derived composite of acceleration changes across three axes, for each player across the five sessions and ran receiver operating characteristic analysis against the injury endpoint. The result was sobering: an area under the curve of just 0.577, barely better than a coin flip. An exploratory analysis of high-intensity running distance, the sum of distances covered in the two fastest velocity bands, fared no better, with an AUC of 0.532. Notably, the workload threshold derived from the data fell numerically within the famous 0.8 to 1.3 acute-to-chronic workload corridor proposed in Gabbett’s training-injury prevention paradox, the so-called sweet spot, and yet it still failed to discriminate who would get hurt.

The biological alternative rested on a different premise: that the same external load can mean very different things to different bodies, and that the blood reveals how an individual organism actually processed a given training stimulus. The team measured four serum cytokines, IL-6, IL-8, IL-10, and TNF-alpha, using a magnetic bead-based multiplex immunoassay on a Luminex platform. Each marker captures a complementary dimension of the acute exercise response: IL-6 and TNF-alpha as central inflammatory mediators, IL-8 as a chemotactic signal governing leukocyte trafficking, and IL-10 as the anti-inflammatory counterweight that reins the response back in.

Alongside the cytokines, the researchers probed the cellular power plants. They isolated peripheral blood mononuclear cells from whole blood using Ficoll density gradient centrifugation and quantified intracellular ATP with a luciferase-based luminescence assay, a readout of each cell’s bioenergetic status. Mitochondrial vitality was approximated with a proximity ligation assay, an elegant microscopy technique that generates a fluorescent signal only when two target proteins sit in close spatial proximity. Here the targets were TFAM and TFB2M, mitochondrial transcription factors whose interaction serves as a surrogate for the integrity and regulatory capacity of the mitochondrion’s own genetic machinery. At least fifty cells per sample were analyzed to ensure robust quantification.

To convert these six biomarkers into a usable clinical tool, the team derived an individual cut-off for each one using ROC analysis with the Youden index, the threshold that best balances sensitivity and specificity. The cut-offs were strikingly concrete: a maximum IL-6 of 1.335 picograms per milliliter, a maximum IL-8 of 27.72 picograms per milliliter, a maximum IL-10 of 1.425 picograms per milliliter, a post-training TNF-alpha of 3.66 picograms per milliliter, a training-induced increase in the TFAM/TFB2M interaction of 19 percent or more, and an ATP increase of 91.5 percent or more. Each exceeded threshold contributed one point, yielding a composite Injury score ranging from zero to six, with every biomarker weighted equally to keep the model transparent and resistant to overfitting in the small cohort.

The performance gap was stark. The biomarker-derived score achieved an apparent AUC of 0.939, with a 95 percent confidence interval of 0.828 to 1.000, indicating high discriminatory capacity for injuries occurring within the next ten days. Using a threshold of four points, players scoring below four had a 93 percent probability of remaining injury-free, thirteen of fourteen athletes, while those scoring four or above went on to suffer an injury in 71 percent of cases, five of seven. The score translated into three clinically interpretable risk bands: zero to one points indicating very low risk, two to three indicating low risk, and four to six indicating increased risk. Internal validation with two thousand bootstrap resamples, in which all cut-offs were re-derived in each resample, corrected the apparent AUC down to 0.870, still a strong figure.

The mechanistic story behind the score is nuanced. Exercise-induced inflammation is not inherently harmful; at physiological doses, reactive oxygen species and cytokines such as IL-6 act as redox-sensitive signaling molecules that drive mitochondrial biogenesis, antioxidant defense, and metabolic adaptation. IL-6, for instance, activates STAT3, AMPK, and p38 MAPK signaling in skeletal muscle and helps coordinate substrate availability and post-exercise recovery. The problem arises when the response overshoots the athlete’s adaptive capacity. Experimentally, TNF-alpha increases oxidative stress and depresses myofibrillar force generation in muscle, and exaggerated systemic inflammatory responses have been linked to impaired neuromuscular recovery and transient tissue vulnerability. The authors are careful to stress that the biomarkers should be read as indicators of systemic stress, not as direct causal mediators of musculoskeletal injury, and that PBMC-derived signals may not mirror local processes in muscle or tendon.

Practical implementation currently requires serious laboratory infrastructure: serum cytokine measurement, PBMC processing, luminescence-based ATP quantification, and microscopy-based proximity ligation assays. The authors argue these are all achievable on routinely available platforms with moderate consumable costs once infrastructure exists, and that future multiplexed assay formats could shrink turnaround times toward something resembling point-of-care use in elite sport. They also emphasize that the findings should not be read as an argument against workload monitoring. Rather, the vision is complementary: external load data interpreted through the lens of each athlete’s individual biological response, potentially extended with additional markers such as IL-1beta, IL-1 receptor antagonist, CCL2/MCP-1, and C-reactive protein.

Significant caveats remain before clubs start drawing blood after every session. The cohort was small, only six injury events occurred, and the score was derived and evaluated in the same group of players, a design that inflates apparent performance even with bootstrap correction. The thresholds may also be specific to this season, this club, and this population, and systemic blood markers cannot capture structural factors such as short biceps femoris fascicles or weak eccentric hamstring strength, which independent research has tied to hamstring injury risk. The authors call for prospective validation in larger, independent cohorts, ideally across different sports and competitive settings. Still, the core message is provocative: the most informative injury warning system may not measure how far a player ran, but how his body responded to the running he did.

Subject of Research: A biomarker-derived injury risk score for professional football players based on inflammatory cytokines and mitochondrial vitality markers

Article Title: A biomarker-derived injury score integrating inflammatory cytokines and mitochondrial vitality may predict short-term injury risk in football

Article References: Ziehe, D., Koos, B., Thon, P., Nowak, H., Hennigs, N., Schepanek, K., Ingelfinger, P., Schlüter, A., Holthusen, H., Meichelbeck, M., Sandfort, M., Gebhardt, K., Stuby, F., Krüger, K., & Adamzik, M. (2026). A biomarker-derived injury score integrating inflammatory cytokines and mitochondrial vitality may predict short-term injury risk in football. Frontiers in Sports and Active Living, 8, Article 1913729. https://doi.org/10.3389/fspor.2026.1913729

Image Credits: AI Generated

DOI: 10.3389/fspor.2026.1913729

Keywords: biomarkers, injury prediction, elite football, cytokines, mitochondrial vitality, ATP, training load, total player load, sports immunology, ROC analysis, injury prevention, Bundesliga