Therefore, it is vital for all of us to treat regulatory actions and postmarketing adverse event reports seriously. When a critical mass of certain adverse events or deaths emerges, the burden lies on the manufacturer to prove a drug does not cause this problem. The burden does not lie with patients, clinicians, or regulatory agencies to prove a drug is unsafe. One prospective naturalistic study has suggested that approximately 36 percent of patients taking clozapine over a five-year period may be diagnosed with diabetes (Am J Psychiatry, 2000; 157: 975-981), a rate of about 7 percent a year. This rate is several-fold higher than that reported by other studies with shorter drug-exposure periods. Currently, no one knows what the true rates of treatment-emergent diabetes are with any atypical antipsychotic, and causality is not proven. However, if a drug with a million patient-years of exposure were to precipitate diabetes at a rate of 3 percent to 7 percent a year, then there would be an estimated 30,000-70,000 cases of treatment-emergent diabetes mellitus. Because atypical antipsychotics will be taken by millions of individuals for decades, large pragmatic trials conducted by neutral parties to address causality and relative risk differences are essential.