Candidates for a new drug to treat a disease might theoretically include from 5,000 to 10,000 chemical compounds. On average about 250 of these will show sufficient promise for further evaluation using laboratory tests, mice and other test animals. Typically, about ten of these will qualify for tests on humans. A study conducted by the Tufts Center for the Study of Drug Development covering the 1980s and 1990s found that only 21.5 percent of drugs that start phase I trials are eventually approved for marketing. The high failure rates associated with pharmaceutical development are referred to as the «attrition rate» problem. Careful decision making during drug development is essential to avoid costly failures. In many cases, intelligent programme and clinical trial design can prevent false negative results. Well designed dose-finding studies and comparisons against both a placebo and a gold-standard treatment arm play a major role in achieving reliable data.