Nineteenth and 20th century contributions to the case-control study are briefly described. Case-control and case-cohort designs are often used today to limit ascertainment of expensive biomarker and genomic data to the most informative participants in a cohort study. Published reports, however, are usually based only on complete data records for cases and controls and thus ignore substantial amounts of information collected for the remaining members of the cohort. The Atherosclerosis Risk in Communities (ARIC) investigators, for example, typically reported data only for the 10-15% of subjects sampled for sub-studies of their cohort of 15,972 participants. The remaining subjects contributed to stratified sampling weights, but not otherwise. Horwitz-Thompson estimation in semiparametric models, with sampling weights adjusted to improve estimation effiency, offers a robust and widely applicable approach to recovery of this information. Recent implementation of this approach in standard computer packages means that biostatisticians and epidemiologists need no longer waste valuable information. In re-analysis of data from an ARIC case-cohort study, no improvement was found for hazard ratios linking lipoprotein-associated phospholipase A2 (Lp-PLA2) with risk of coronary heart disease. The standard error of the interaction of Lp-PLA2 with systolic blood pressure was reduced by 10%, however, and the precision of hazard ratios for covariates used for adjustment improved dramatically. Many other studies would similarly benefit from routine application of these methods.