Painful crises in patients with sickle cell anemia are caused by vaso-occlusion and infarction. Occlusion of blood vessels depends on (at least) their diameter, the deformability of red cells, and the adhesion of blood cells to endothelium. Deoxygenated sickle cells are rigid because they contain linear polymers of hemoglobin S (Hb S); polymerization is highly concentration dependent, and dilution of Hb S by a nonsickling hemoglobin such as fetal hemoglobin (Hb F) would be expected to lead ultimately to a decrease in the frequency of painful crises. It might also be expected to decrease the severity of anemia, although the pathogenesis of anemia in sickle cell anemia (SS disease) is not clearly understood. Reversion to production of fetal rather than adult hemoglobin became practical with the discovery that HU was an orally effective and relatively safe "switching agent." Preliminary dose-ranging studies led to a double-blind randomized controlled clinical trial, the Multicenter Study of Hydroxyurea in Sickle Cell Anemia (MSH), designed to test whether patients treated with HU would have fewer crises than patients treated with placebo. The MSH was not designed to assess the mechanism(s) by which a beneficial effect might be achieved, but it was hoped that observations made during the study might illuminate that question. The 2 MSH treatment groups were similar to each other and were representative of African-American patients with relatively severe disease. The trial was closed earlier than expected, after demonstration that median crisis rate was reduced by almost 50% (2.5 versus 4.5 crises per year) in patients assigned to HU therapy. Hospitalizations, episodes of chest syndrome, and numbers of transfusions were also lower in patients treated with HU. Eight patients died during the trial, and treatment was stopped in 53. There were no instances of alarming toxicity. Patients varied widely in their maximum tolerated doses, but it was not clear that all were taking their prescribed treatments. When crisis frequency was compared with various clinical and laboratory measurements, pretreatment crisis rate and treatment with HU were clearly related to crisis rate during treatment. Pretreatment laboratory measurements were not associated with crisis rates during the study in either treatment group. It was not clear that clinical improvement was associated with an increase in Hb F. Crisis rates of the 2 treatment groups became different within 3 months. Mean corpuscular volumes (MCVs) and the proportion of Hb F containing red cells (F cells) rose, and neutrophil and reticulocyte counts fell, within 7 weeks. When patients were compared on the basis of 2-year crisis rates, those with lower crisis rates had higher F-cell counts and MCVs and lower neutrophil counts. Neutrophil, monocyte, reticulocyte, and platelet counts were directly associated, and F cells and MCV were inversely associated, with crisis rates in 3-month periods. In multivariable analyses, there was strong evidence of independent association of lower neutrophil counts with lower crisis rates. F-cell counts were associated with crisis rate only in the first 3 months of treatment; MCV showed an association over longer periods of time. Overall, the evidence that decreased neutrophil counts played a role in reducing crisis rates was strong. Increased F cells or MCV and evidence of cytoreduction by HU were also associated with decreased crisis rates, but no definitive statement can be made regarding the mechanism of action of HU because the study was not designed to address that question. Future studies should be designed to explore the mechanism of action of HU, to identify the optimal dosage regimen, and to study the effect of HU when combined with other antisickling agents.