To evaluate the effectiveness of three models for pre-exposure prophylaxis (PrEP) service delivery to HIV-1 serodiscordant couples in Nigeria. 297 heterosexual HIV-1 serodiscordant couples were recruited into three PrEP delivery models and followed up for 18 months. The models were i) Outpatient clinic model providing PreP in routine outpatient care; ii) Antiretroviral therapy (ART) clinic model providing PrEP in ART clinics; and iii) Decentralized care model providing PrEP through primary and secondary care centres linked to a tertiary care centre. The primary effectiveness endpoint was incident HIV-1 infection. The HIV incidence before and after the study was compared and the incidence rate ratio computed for each model. Survival analysis was conducted, Cox regression analysis was used to compare the factors that influenced couple retention in each of the models. Kaplan-Meier survival analysis was used to estimate the median retention time (in months) of the study participants in each of the study models, and log-rank test for equality of survival functions was conducted to test for significant differences among the three models. There was no significant difference (p>0.05) in the couple retention rates among the three models. At months 3, 6 and 9, adherence of the HIV-1-infected partners to ART was highest in the decentralized model, whereas at months 9 and 12, the outpatient model had the highest proportion of HIV-1- uninfected partners adhering to PrEP (p<0.001). The HIV incidence per 100 person-years was zero in the general outpatient clinic and ART clinic models and 1.6 (95% CI: 0.04-9.1) in the decentralized clinic model. The difference in the observed and expected incidence rate was 4.3 (95% CI: 0.44-39.57) for the decentralized clinic model. Although incidence of HIV seroconversion was highest in the decentralized clinic model, this difference may be due to the higher sexual risk behavior among study participants in the decentralized model rather than the type of service delivery. The study findings imply that any of the models can effectively deliver PrEP services.