Statistical models used to forecast malignant pleural mesothelioma (MPM) trends often do not take into account historical asbestos consumption, possibly resulting in less accurate predictions of the future MPM death toll. We used the distributed lag non-linear model (DLNM) approach to predict future MPM cases in Italy until 2040, based on past asbestos consumption figures. Analyses were conducted using data on male MPM deaths (1970-2014) and annual asbestos consumption using data on domestic production, importation, and exportation. According to our model, the peak of MPM deaths is expected to occur in 2021 (1122 expected cases), with a subsequent decrease in mortality (344 MPM deaths in 2039). The exposure-response curve shows that relative risk (RR) of MPM increased almost linearly for lower levels of exposure but flattened at higher levels. The lag-specific RR grew until 30 years since exposure and decreased thereafter, suggesting that the most relevant contributions to the risk come from exposures which occurred 20-40 years before death. Our results show that the Italian MPM epidemic is approaching its peak and underline that the association between temporal trends of MPM and time since exposure to asbestos is not monotonic, suggesting a lesser role of remote exposures in the development of MPM than previously assumed.