Author(s): BANDIERA, D., GARRANDES, F., MATZARAKIS, A., ADAMI, P.E., RACINAIS, S., PITSILADIS, Y., TESSITORE, A., BERMON, S., Institution: UNIVERSITY OF ROME FORO ITALICO, Country: ITALY, Abstract-ID: 2277

INTRODUCTION:
Exercising under strenuous environmental conditions can lead to fatal exertional heat stroke (EHS) if not managed promptly and adequately. Also, the number of athletes who do not finish a race (DNF) is believed to be influenced by environmental stress during competition. To measure heat stress, international federations rely on thermal indices, such as the Wet Bulb Globe Temperature (WBGT) in Athletics. However, the relationship between thermal indices, EHS incidence, and DNF rates in elite endurance athletes remains unclear. The aim of this study was therefore to i) quantify the incidence of EHS and DNF events in World Athletics competitions, ii) examine their association with thermal indices, and iii) develop predictive models to estimate EHS and DNF risk before a race.
METHODS:
Eighty Athletics endurance races held between 2019 and 2024 were included in the analysis, encompassing the World Championships and the Olympic Games. Environmental parameters (e.g., temperature, humidity) and WBGT were measured on-site and linked to race-specific details (e.g., distance, race time), and the number of DNF and EHS (called dependent variables) were counted. Additionally, several thermal indices (e.g., PET, mPET, UTCI) and energy expenditure were estimated for each race. To analyse the relationship between predictors and the dependent variables, a linear regression analysis was conducted, with a significance level at 0.05. Additionally, Generalized Additive Models (GAM) with k-fold cross validation were used to assess the prediction power of models on unknown races.
RESULTS:
A total of 41 EHS cases were diagnosed among 4938 athletes (8/1000 athlete exposures) with 56 races (70%) reporting no EHS and 24 races (30%) recording between 1 and 5 cases. In average, each race led 6 ± 9 DNF, ranging from 0 to 60. The strongest EHS linear regression was observed for UTCI (R²=0.10, p=0.003), while WBGT showed a lower association (R²=0.06, p=0.025). For DNF, race distance was the strongest linear regression (R²=0.40, p<0.001) while WBGT association was not significant (R²=0.01, p=0.58). The best GAM for predicting EHS included temperature, WBGT, mean radiant temperature, and the days elapsed since the last spring onset, with an R² of 0.20 ± 0.15. For DNF, the best GAM included WBGT, race time, and mean velocity, achieving an R² of 0.80 ± 0.03.
CONCLUSION:
Our analysis showed that WBGT alone has only a weak linear relationship with EHS and DNF in Athletics and that more recent thermal indices (e.g., UTCI), while improved, still do not demonstrate a sufficiently strong association. Likewise, non-linear prediction models incorporating environmental parameters and race-specific details do not provide a robust estimation of EHS. This suggests that key factors contributing to EHS were not accounted for in our analysis. However, the prediction models for DNF proved to be robust, allowing for an accurate estimation of the number of athletes who will not finish the race, before the start of the competition.