We reviewed the cases of thirty-four patients (thirty-eight knees) in whom prostheses with an identical design were used for primary total knee arthroplasty, and we compared the perioperative blood loss between the eleven knees in which cement was used and the twenty-seven knees in which cement was not used. The patients who had an uncemented prosthesis had a significantly greater mean blood loss, both intraoperatively (p less than or equal to 0.05) and during each subsequent eight-hour interval on the first postoperative day (p less than or equal to 0.05). The total for the forty-eight-hour postoperative collection also was greater (p less than or equal to 0.01), as was the cumulative loss for the entire study (p less than or equal to 0.01). When patients who had rheumatoid arthritis and osteoarthrosis were considered separately, the results were similar; that is, there was a significantly greater total postoperative blood loss in each of the two groups when cement was not used (p less than or equal to 0.025). A minimum postoperative hemoglobin concentration of ninety-five grams per liter was maintained; a greater percentage of patients (sixteen of twenty-seven) in whom cement was not used needed a transfusion as compared with two of eleven in the group in whom cement was not used (p less than or equal to 0.025), and they also needed more packed red-blood cells (1062 compared with 750 milliliters) (p less than or equal to 0.05).(ABSTRACT TRUNCATED AT 250 WORDS)