Abstract In the U.S., Latina women report higher rates of inactivity than their non-Hispanic White and male counterparts, and are disproportionately affected by related health conditions (e.g., cancer, hypertension, heart disease, stroke, diabetes). To address this public health crisis, evidence-based interventions that utilize state- of-the-art technology, theory and methods are needed to increase physical activity (PA) among this high-risk population. Recently, our team conducted a randomized controlled trial (N=205) to test the efficacy of a culturally adapted, individually tailored, Spanish-language Internet-based PA intervention among Latinas (Pasos Hacia La Salud, R01CA159954) vs. a Wellness Contact Control Internet Group. Although results were promising, the majority of participants still did not meet national guidelines for PA. In the ongoing renewal of R01CA159954, we will randomize 300 Latina women to either 1) the original Pasos Hacia La Salud tailored Internet-based PA intervention (Original Intervention) or 2) the data driven, enhanced version of the Pasos Hacia La Salud PA intervention (Enhanced Intervention). The proposed supplement seeks to become the first trial to examine longitudinal patterns of PA adoption and maintenance across a series of longitudinal studies among Latinas. Traditional analytic methods would compare findings across studies and would estimate intervention effects at end of treatment controlling for baseline. We propose to use Integrative Data Analysis to combine data across studies and Latent Class Models to identifies patterns of PA adoption and maintenance. By using Latent Class Modeling (LCM), we will make full use of the longitudinal profile of PA data to identify patterns of change over time. LCMs use objective data (rather than a priori hypotheses) to subdivide the population into distinct groups of participants with similar profiles (of PA changes in this case). LCM can not only capture between- and within-subject heterogeneity, but can aid in the identification of meaningful subgroups within the population. By using this innovative method, we will not only reveal the inter-variability of PA over time but also elucidate the associations between such patterns and key psychosocial, behavioral and demographic predictors.