The scientific and public health implications of gene-environment interaction warrant that the most powerful study designs and methods of analysis be used. Because traditional case-control designs, which use nonrelated subjects, have demonstrated the need for large samples to detect interactions, alternative study designs may be worthwhile, such as sampling diseased cases and their parents. If the transmission of particular alleles from parents to their diseased child appears to be distorted from Mendelian expectation, then this suggests an etiologic association of the alleles with disease; if the frequency of transmission differs between exposed and nonexposed cases, then gene-environment interaction is suggested. We present likelihood-based methods to assess interaction, as well as an extension of the transmission/disequilibrium test (TDT). For these statistical tests, we also derive methods to compute sample size and power. Comparisons of sample size requirements between the case-parents design and the case-control design indicate that the case-parents design can be more powerful to detect gene-environment interactions, particularly when the disease susceptible allele is rare. Also, one of the derived likelihood methods, based on additive effects of alleles, tended to be the most robust in terms of power for a broad range of genetic mechanisms, and so may be useful for broad applications to assess gene-environment interactions.