To study the distribution rule of modified time-to-collision (MTTC) in vehicle-to-vehicle (V2V) communications, a mixed distribution model that can describe the MTTC distribution of different driving risk levels was proposed, and the model was verified by field test data. Experiment on V2V communications was carried out based on long-term evolution-vehicle (LTE-V) technology. Real vehicle driving test data was obtained, and k-means clustering method was used to divide the driving risk levels into four levels. Then, three probability distribution models including Weibull, Gamma, and lognormal were utilized, and the goodness-of-fit analysis was carried out for MTTC under four risk levels. Test results show that MTTC under different risk levels could be fitted well by lognormal distribution model. On this basis, an MTTC lognormal mixed distribution model for four risk levels was established, and the model parameters were calibrated using expectation-maximization (EM) algorithm. Three classic distribution models (Weibull, Gamma, and lognormal) were selected for comparison. The effectiveness of the proposed mixed distribution model was verified by goodness-of-fit analysis, and the mixed distribution model was applied to MTTC test data in conventional environment. Results indicate that the proposed lognormal mixed distribution model can fit the MTTC distribution better in V2V communication environment and has good adaptability to the conventional environment, which can provide theoretical support for related research on vehicle operation safety.