Diagnosis of pyruvate kinase deficiency (PKD), the most common cause of hereditary non-spherocytic haemolytic anaemia, remains challenging in routine practice and no biomarkers for clinical severity have been characterised. This prospective study enrolled 41 patients with molecularly confirmed PKD from nine North American centres to evaluate the diagnostic sensitivity of pyruvate kinase (PK) enzyme activity and PK:hexokinase (HK) enzyme activity ratio, and evaluate the erythrocyte PK (PK-R) protein level and erythrocyte metabolites as biomarkers for clinical severity. In this population not transfused for ≥90 days before sampling, the diagnostic sensitivity of the PK enzyme assay was 90% [95% confidence interval (CI) 77-97%], whereas the PK:HK ratio sensitivity was 98% (95% CI 87-100%). There was no correlation between PK enzyme activity and clinical severity. Transfusion requirements correlated with normalised erythrocyte ATP levels (r = 0·527, P = 0·0016) and PK-R protein levels (r = -0·527, P = 0·0028). PK-R protein levels were significantly higher in the never transfused [median (range) 40·1 (9·8-73·9)%] versus ever transfused [median (range) 7·7 (0·4-15·1)%] patients (P = 0·0014). The PK:HK ratio had excellent sensitivity for PK diagnosis, superior to PKLR exon sequencing. Given that the number of PKLR variants and genotype combinations limits prognostication based on molecular findings, PK-R protein level may be a useful prognostic biomarker of disease severity and merits further study.