Background The Para athletics wheelchair-racing classification system employs best practice to ensure that classes comprise athletes whose impairments cause a comparable degree of activity limitation. However, decision-making is largely subjective and scientific evidence which reduces this subjectivity is required.

Aim To evaluate whether isometric strength tests were valid for the purposes of classifying wheelchair racers and whether cluster analysis of the strength measures produced a valid classification structure.

Methods Thirty-two international level, male wheelchair racers from classes T51–54 completed six isometric strength tests evaluating elbow extensors, shoulder flexors, trunk flexors and forearm pronators and two wheelchair performance tests–Top-Speed (0–15 m) and Top-Speed (absolute). Strength tests significantly correlated with wheelchair performance were included in a cluster analysis and the validity of the resulting clusters was assessed.

Results All six strength tests correlated with performance (r=0.54–0.88). Cluster analysis yielded four clusters with reasonable overall structure (mean silhouette coefficient=0.58) and large intercluster strength differences. Six athletes (19%) were allocated to clusters that did not align with their current class. While the mean wheelchair racing performance of the resulting clusters was unequivocally hierarchical, the mean performance of current classes was not, with no difference between current classes T53 and T54.

Conclusions Cluster analysis of isometric strength tests produced classes comprising athletes who experienced a similar degree of activity limitation. The strength tests reported can provide the basis for a new, more transparent, less subjective wheelchair racing classification system, pending replication of these findings in a larger, representative sample. This paper also provides guidance for development of evidence-based systems in other Para sports.