Drugs lost in translation in CKD: from redundant designs to learning systems.

Chronic kidney disease (CKD) remains a leading cause of morbidity and mortality, yet many promising drugs fail in late-stage development despite plausible biology and strong financial investment. This failure often reflects redundancy in the translational pathway rather than intrinsic lack of efficacy. Preclinical research continues to rely on narrow, reductionist models that poorly recapitulate the chronic, multifactorial nature of human CKD and its cardiovascular complications. Early-phase clinical trials frequently prioritise short-term surrogate endpoints and broad, heterogeneous CKD populations, diluting biologically coherent subgroups and obscuring clinically meaningful effects. Late-phase programmes, exemplified by the recent ocedurenone experience, are then terminated for neutral primary outcomes or safety concerns, with limited transparency and learning from negative data. We argue that CKD drug development must shift towards learning health systems built on human-relevant models, biology-driven enrichment, more appropriate and patient-centred endpoints, and adaptive trial platforms. Systematic reporting and shared analysis of failed and futile trials are essential to avoid repeating avoidable errors. Re-engineering this translational ecosystem is crucial to convert mechanistic insight into durable therapeutic advances for people living with CKD.