Purpose We aimed to investigate biomarkers and predictive factors for visual and anatomical outcome in patients with naïve diabetic macular edema (DME) who underwent small gauge pars plana vitrectomy (PPV) with internal limiting membrane (ILM) peeling as a first line treatment. Design Multicenter, retrospective, interventional study. Participants 120 eyes from 120 patients with naïve DME treated with PPV and ILM peeling with a follow up of 24 months. Methods Change in baseline best corrected visual acuity (BCVA) and central subfoveal thickness (CST) 1, 6, 12 and 24 months after surgery. Predictive value of baseline BCVA, CST, optical coherence tomography (OCT) features (presence of subretinal fluid (SRF) and photoreceptor damage), and time between DME diagnosis and surgery. Additional treatment for DME needed. Intra- and post-operative complications (cataract rate formation, increased intraocular pressure). Main outcome measures The correlation between baseline characteristics and BCVA response (mean change from baseline; categorized improvement 5 or 10; Early Treatment Diabetic Retinopathy Study (ETDRS) letters) 12 and 24 months after surgery. Results Mean BCVA was 0.66 ± 0.14 logMAR, 0.52 ± 0.21 logMAR, and 0.53 ± 0.21 logMAR (p<0.001) at baseline, 12 and 24 months, respectively. Shorter time from DME diagnosis until PPV (OR: 0.98, 95% CI: 0.97–0.99, p<0.001) was a predictor for good functional treatment response (area under the curve 0.828). For every day PPV is postponed, the patient’s chances to gain 5 letters at 24 months decrease by 1.8%. Presence of SRF was identified as an anatomical predictor of a better visual outcome, (OR: 6.29, 95% CI: 1.16–34.08, p = 0.033). Safety profile was acceptable. Conclusions Our results reveal a significant functional and anatomical improvement of DME 24 months after primary PPV, without the need for additional treatment. Early surgical intervention and presence of SRF predict good visual outcome. These biomarkers should be considered when treatment is chosen.