In France, as in other countries, accident research studies show that the greatest proportion of restrained occupants sustaining severe injuries and fatalities are involved in frontal impact (70% and 50% respectively). In severe frontal impacts with restraint occupants and where intrusion is not preponderant, the oldest occupants very often sustain severe thoracic injuries due to the seat belt. In the seventies, a few cars were equipped in France with load limiters and it was thereby possible to observe a relationship between the force applied and the occupant's age with regard to this thoracic risk.The reduction of intrusion for the most violent frontal impacts, through optimization of car deformation, usually translates into an increase in restraint forces and hence thoracic risks with a conventional retractor seat belt for a given impact violence. It is therefore essential to limit restraint forces with a seat belt to reduce the number of road casualties, especially for the most elderly.In order to address the thoracic risk in frontal impact, a restraint system combining belt load limitation and pyrotechnic belt pretension, known as the Programmed Restraint System (PRS), has been fitted to Renault cars since 1995. The belt load limiter, which is a steel part designed to deform at a given shoulder force threshold, i.e. 6 kN, is fastened between the retractor and the lower anchorage point of the belt.From static and dynamic tests performed with the load limiter, it is possible to determine the shoulder belt force applied to the occupant from the amount of load limiter deformation. Eighty-nine accident cases, with EES's ranging (Equivalent Energy Speed) from 40 to 80 km/h and involving frontal collisions with cars equipped with the PRS, are reported in this paper.The purpose of the present study, based on this accident file and also on another accident file with textile limiters, is to establish, for belted occupants, the thoracic injury risk as a function of occupant age and the load applied at shoulder level. One observes that for 50% of thorax risk of AIS3+, the force for all ages together is 6.9 kN. Occupant's age is a very important factor. These results are obtained for 256 occupants, where the age's distribution is similar to that of front seat occupants of the French accident file. Despite occupant's sizes and belt geometries in this accident sample, the shoulder belt load appears to be in accordance with the occurrence of chest injuries.A comparison of results obtained in real-world accidents and the tests performed with 209 PMHS on Laboratory (Post Mortem Human Subject) by Heidelberg University is also described in this paper. Both databases show the same trend for the thoracic risk according to age, but the force for a given risk and a given age is 2 kN lower than that for occupants in real-world accidents.A relationship between the Hybrid III thoracic injury measurements and the shoulder belt load is also investigated in this paper. For a given shoulder belt load in crash tests performed in cars identical to those in the accident analyzed with PRS, the sternal deflection scatter does not allow to satisfy evaluation of the thoracic risk. These results are discussed and compared with several papers. It is suggested to include the shoulder belt load measurement in homologation and rating tests for a better assessment of thorax protection.This study confirms that a 6 kN force level is not sufficient to protect a larger proportion of the population. A belt load limitation of 4 kN, combined with a specifically design airbag, would make it possible to protect 95% of those involved in frontal impacts from thorax injuries of AIS3+.