A form of deception of subjects can occur if the terms of the informed consent are violated by the investigator or other scientists. For example, if a subject consents to have their blood sample evaluated for markers of diabetes and the investigator or other scientists use the blood sample for purposes unrelated to diabetes research, the subject has effectively been deceived about the nature of their research participation and they have been denied autonomous agency over their own actions. The recent court case involving Arizona State University v. the Havasupai tribe illustrates the ethical pitfalls and legal consequences of biological sample sharing without explicit prior consent (Drabiak-Syed, 2010).
The concept of non-maleficence is embodied by the phrase, "first, do no harm," or the Latin, primum non nocere . Many consider that should be the main or primary consideration (hence primum ): that it is more important not to harm your patient, than to do them good, which is part of the hippocratic oath that doctors take.  This is partly because enthusiastic practitioners are prone to using treatments that they believe will do good, without first having evaluated them adequately to ensure they do no harm to the patient. Much harm has been done to patients as a result, as in the saying, "The treatment was a success, but the patient died." It is not only more important to do no harm than to do good; it is also important to know how likely it is that your treatment will harm a patient. So a physician should go further than not prescribing medications they know to be harmful—he or she should not prescribe medications (or otherwise treat the patient) unless s/he knows that the treatment is unlikely to be harmful; or at the very least, that patient understands the risks and benefits, and that the likely benefits outweigh the likely risks.