Although the number of cases evaluated by Girard and colleagues1 was small, their results illustrate the need to incorporate molecular information into our categorization scheme for multiple lung neoplasms. Another recently published study using loss of heterozygosity studies, TP53 mutation screening analyses, and evaluations of X-chromosome inactivation patterns to assess patients with multiple lung tumors further highlights the utility of molecular methods for offering insight into this question.4 Larger-scale integration of molecular data into clinical decision making was anticipated by the authors of the ACCP guidelines,3 and it is inevitable that future classification systems will include greater representation of molecular information as we seek to divine biologic truth. What do we ask of a successful classification system? It should provide a common language to facilitate clinical care and research. It must demonstrate power for predicting biologic behavior that is validated against outcomes. The techniques used to generate the data needed for classification should ideally be widely available and straightforward to apply, yielding data elements with high intraobserver and interobserver reliability. Low cost and rapid turnaround time would likewise be advantageous. Tests should deliver highly sensitive and specific results that are not greatly affected by tumor preservation or heterogeneity. Additionally, in the best of all worlds, the classification system can supply information helpful for devising a successful treatment plan (ie, predict responsiveness to specific therapies) and can be applied to small samples obtained using minimally invasive procedures. In the next iterations of criteria for differentiating between multiple primary carcinomas and metastases, we should prepare for a transition in the types of data elements we consider, as the utility or redundancy of macroscopic, microscopic, molecular, and temporal factors is reevaluated.