Purpose To determine the association between serum periostin and the presence of diabetic retinopathy (DR). Methods Serum periostin was detected in 114 healthy subjects, 122 patients with type 2 diabetes mellitus (T2DM) and 159 patients with DR and compared among groups. Clinical data and other laboratory measurements such as glycated hemoglobin (HbA1c), lipid profiles, serum creatinine (Cr) and high-sensitivity CRP (hsCRP) were also collected and compared among groups. For subgroup analysis, patients with DR were divided into a non-proliferated diabetic retinopathy (NPDR) group and a proliferated diabetic retinopathy (PDR) group. Multivariate analysis was performed using logistic regression models. Results The serum periostin level was significantly higher in patients with diabetic retinopathy compared with healthy subjects and patients with T2DM (both P  < 0.001, respectively). Also, the periostin level was significantly higher in the PDR group compared to the NPDR group ( P  = 0.044). Multivariate logistic regression revealed that serum periostin was independently associated with the presence of DR in patients with T2DM ( P  < 0.001). The receiver operating characteristic (ROC) curves for DR development using serum periostin showed that the area under the receiver operating characteristic curves (AUC) was 0.838 ( P  < 0.001). Conclusions The current study demonstrated that serum periostin is significantly associated with the presence of DR in patients with T2DM and is an independent risk factor of DR.