Background. The utility of renal biopsy in patients with diabetes is highly debated. Diabetics with rapidly worsening renal disease are often 'clinically' labelled as having diabetic nephropathy (DN), whereas, in many cases, they are rather developing a nondiabetic renal disease (NDRD) or mixed forms (DN þ NDRD). Methods. We performed a systematic search for studies on patients with diabetes with data on the frequency of DN, NDRD and mixed forms, and assessed the positive predictive values (PPVs) and odds ratios (ORs) for such diagnoses by metaanalysing single-study prevalence. Possible factors explaining heterogeneity among the different diagnoses were explored by meta-regression. Results. In the 48 included studies (n ¼ 4876), the prevalence of DN, NDRD and mixed forms ranged from 6.5 to 94%, 3 to 82.9% and 4 to 45.5% of the overall diagnoses, respectively. IgA nephropathy was the most common NDRD (3-59%). PPVs for DN, NDRD and mixed forms were 50.1% [95% confidence interval (CI): 44.7-55.2], 36.9% (95% CI: 32.3-41.8) and 19.7% (95% CI: 16.3-23.6), respectively. The PPV when combining NDRD and mixed forms was 49.2% (95% CI: 43.8-54.5). Metaregression identified systolic pressure, HbA1c, diabetes duration and retinopathy as factors explaining heterogeneity for NDRD, creatinine and glomerular filtration rate for mixed forms and only serum creatinine for DN. ORs of DN versus NDRD and mixed forms were 1.71 (95% CI: 1.54-1.91) and 4.1 (95% CI: 3.43-4.80), respectively. Conclusions. NDRD are highly prevalent in patients with diabetes. Clinical judgment alone can lead to wrong diagnoses and delay the establishment of adequate therapies. Risk stratification according to individual factors is needed for selecting patients who might benefit from biopsy.