Purpose: To ascertain whether in the presence of a previously endodontically treated tooth with periapical pathology and/or symptoms and an uncertain prognosis, it is better to endodontically retreat it or to replace the tooth with a single-implant-supported crown. Materials and methods: Twenty patients requiring the treatment of a previously endodontically treated tooth, with periapical pathology and/or symptoms of endodontic origin and an uncertain prognosis, as judged by the recruiting investigator, were randomly allocated to endodontic retreatment (Endo group, 10 patients) or tooth extraction and replacement with an implant-supported crown (Implant group, 10 patients) according to a parallel-group design at a single centre. Patients were followed to 3 years after completion of the treatment. Outcome measures were: failure of the procedure, complications, marginal bone level changes at both teeth and implants, endodontic radiographic success (teeth only), number of patients' visits and days to complete the treatment, patients' chair time, costs, aesthetics assessed using the pink aesthetic score (PES) for the soft tissues and the white aesthetic score (WES) for the tooth/crown recorded by independent assessors. Results: No patient dropped out. One endodontically retreated tooth fractured and another had a crown loosening. There were no statistically significant differences for treatment failure or complications (difference in proportions = 0.10; 95% CI: -0.09 to 0.29; P (Fisher exact test) = 1.000). The mean marginal bone levels at endodontic retreatment/implant insertion were 2.10 ± 0.66 mm for the Endo group and 0.05 ± 0.15 mm for the Implant group. Three years after completion of the treatment, teeth lost on average 0.23 ± 0.82 mm and implants 0.62 ± 0.68 mm, the difference not being statistically significant (mean difference = -0.39 mm; 95% confidence interval [CI]: -1.12 to 0.33; P (t test) = 0.267). Three years after completion of the endodontic retreatment, of the four teeth that originally had a periapical radiolucency, one was lost, two showed complete healing, and one showed radiographic improvement. There were no statistically significant differences for the number of patients' visits (Endo = 6.67 ± 0.71; Implant = 6.10 ± 0.74; mean difference = 0.57; 95% CI: -0.14 to 1.27; P (t test) = 0.106). It took significantly more days to complete the implant rehabilitation (Endo = 61 ± 12.97; Implant = 191.40 ± 75.04; mean difference = -130.40; 95% CI: -184.45 to -76.35; P (t test) < 0.001) but less patients' chair time (Endo = 629.44 ± 43.62 min; Implant = 326 ± 196.99 min; mean difference = 303.44; 95% CI: 160.87 to 446.02; P (t test) = 0.001). Implant treatment was significantly more expensive (Endo = €1,588.89 ± 300.81; Implant = €2,095 ± 158.90; mean difference = €-506.11; 95% CI: -735.41 to -276.82; P (t test) < 0.001). Three years after treatment completion, mean PES were 11.11 ± 1.97 and 6.50 ± 2.46 and mean WES were 7.78 ± 1.30 and 6.80 ± 2.39 in the Endo group and Implant group, respectively. Soft tissues aesthetics (PES) were significantly better at endodontically retreated teeth (mean difference 4.61; 95% CI: 2.44 to 6.78; P (t test) < 0.001) whereas no significant differences were observed for tooth aesthetics (WES) (mean difference 0.98; 95% CI: -0.89 to 2.85; P (t test) = 0.281) between treatments. Conclusions: The present preliminary results suggest that both endodontic retreatment and replacement of previously endodontically treated teeth with persisting pathology and a dubious endodontic prognosis provided similar short-term success rates. Aesthetics of the soft tissues and time needed to complete treatment were in favour of endodontic retreatment whereas implant rehabilitation required half of the chair time than endodontic retreatment, but was significantly more expensive. Much larger patient populations and longer follow-ups are needed to fully answer this question; however, in this scenario the less invasive endodontic retreatment could be the first therapeutic option to be considered.