Failure of biomarkers to personalize clinical practise in diabetic kidney disease. Are we asking the right questions?
Authors: Sara Denicolò, Felix Keller, Stefanie Thöni, GertMayer (Department of Internal Medicine IV (Nephrology and Hypertension), Medical University Innsbruck), Bernd Mayer (emergentec biodevelopment GmbH, Vienna)
Individuals with type 2 diabetes mellitus (T2DM) and impaired renal function are not homogeneous in their clinical presentation and pathophysiology. Other, non-diabetic kidney diseases (such as glomerulonephritides) may be present, especially if the clinical presentation is atypical (e.g. nephrotic syndrome, haematuria). Patients with a phenotype considered “typical” rarely undergo biopsy, and therefore the prevalence of well-defined non-diabetic pathologies in this population remains unknown. Clinically, progression can take place via the “classical” pattern of “diabetic nephropathy” (DN) going from hyperfiltration to microalbuminuria and macroalbuminuria before eGFR declines,while other cases experience a decrease of eGFR without ever developing proteinuria. To cover this diversity “diabetic kidney disease”(DKD) is the preferred term with DN only being one of many possible disease manifestations. Histological lesions in DKD are also highly variable. Even in cohorts with a relatively homogenous and typical phenotype (microalbuminuria with preserved eGFR), only one third of patients shows biopsy findings indicative of DN (mesangial expansion, thickening of the glomerular basement membrane), while the majority has non specific changes or even normal images on light microscope (Fioretto P; Diabetologia1996; 39:1569). Consequently, the individual prognosis as well as the response to a specific therapy is variable. Inhibition of the renin–angiotensin system, GLP-1 agonist or SGLT2 inhibitor therapy significantly reducedhard endpoints in prospective randomized controlled interventional trials and the implementation of this “cohort-oriented medicine” has improved the prognosis of patients with DKD. However, the “number needed to treat” is high and, when prescribing drugs with side effects, we harm those, who do not adequately respond. In addition, with more effective agents available, the questions about optimum combination therapy is emerging.
Over the last decades, great efforts to better characterize patients with DKD were undertaken to improve the prediction of individual prognosis and treatment response. While prognostic biomarkers aim to predict the trajectory of DKD based on the present (“as-is”) state, predictive biomarkers estimate the consequence of a specific intervention on the prognosis. The European Union has supported several multinational research projects (e.g. SYSKID, SUMMIT, Beat-DKD) to implement stratified or even personalized therapy in complex diseases such as DKD, but unfortunately no novel biomarker candidates have yet found their way into clinical practice. Critics argue that biomarkers in diseases such as DKD are obviously of little value while others propose that extended profiling and application of modern analysis strategies like machine learning will ultimately identify the “right” biomarkers or biomarker panels. The authors of this article support the idea that only biomarkers can lead to“personalized”or “targeted therapy” – and we do see advances in DKD as well. Nonetheless, we critically question, whether it is enough to increase the scale of high throughput screeningas all analysis strategies rely on a specific model of the disease. We believe that it is urgently necessary to re-evaluate our basic concept of progression of DKD to avoid failures that are not based on the quality of the biomarkers chosen. (…)