Although morally acceptable in theory, preimplantation genetic diagnosis (PGD) for mitochondrial DNA (mtDNA) disorders raises several ethical questions in clinical practice. This paper discusses the major conditions for good clinical practice. Our starting point is that PGD for mtDNA mutations should as far as possible be embedded in a scientific research protocol. For every clinical application of PGD for mtDNA disorders, it is not only important to avoid a 'high risk of serious harm' to the future child, but also to consider to what extent it would be possible, desirable and proportional to try to reduce the health risks and minimize harm. The first issue we discuss is oocyte sampling, which may point out whether PGD is feasible for a specific couple. The second issue is whether one blastomere represents the genetic composition of the embryo as a whole -- and how this could (or should) be investigated. The third issue regards the cutoff points below which embryos are considered to be eligible for transfer. We scrutinize how to determine these cutoff points and how to use these cutoff points in clinical practice -- for example, when parents ask to take more or less risks. The fourth issue regards the number of cycles that can (or should) justifiably be carried out to find the best possible embryo. Fifth, we discuss whether follow-up studies should be conducted, particularly the genetic testing of children born after IVF/PGD. Finally, we offer the main information that is required to obtain a truly informed consent.