Given increasing diabetes rates worldwide, better screening tools for diabetic retinopathy (DR) and macular edema (DME) are needed. The study aim was to compare reliability and predictive values between non-mydriatic fundus photography (NMFP) and spectral-domain optical coherence tomography (OCT) for detection of DR and DME with dilated fundus examination (DFE). This was a non-interventional, comparative study. Diabetics underwent both NMFP and macula OCT, followed by DFE. Images were interpreted by two masked ophthalmologists. The DFE result was considered gold standard. One hundred and fifty-four eyes of 83 patients were recruited. Sensitivity of NMFP for DR was 77.3% and 80.3% for OCT. Specificity for NMFP was 81.8% and 55.7% for OCT. Area under Receiver Operating Characteristics Curve (AROC) for DR was 0.80 for NMFP and 0.68 for OCT. The sensitivity of NMFP for DME was 63.2% and 82.5% for OCT. Specificity for DME was 90.1% by NMFP and 61.5% for OCT. Positive predictive value (PPV) of NMFP and OCT for DR was 76.1% (95% CI: 63.9-85.3%) and 57.6% (46.8-67.7%), respectively. Negative predictive value (NPV) of NMFP and OCT was 82.7% (95% CI: 72.8-89.7%) and 79.0% (66.4-87.9%) respectively. Positive predictive value of NMFP and OCT for DME was 80.0% (95% CI: 67.6-88.5%) and 57.3% (45.9-68.0%), respectively. Negative predictive value of NMFP and OCT was 79.6% (95% CI:70.3 - 86.7%) and 84.8% (95% CI:73.4 - 92.1%), respectively. Eyes with normal OCT miss 21% of DR. In conclusion, NMFP is better than OCT for DR screening, while OCT is better than NMFP and DFE for detection of DME. Both modalities should be for better DR screening.