Influence of data display formats on physician investigators' decisions to stop clinical trials: prospective trial with repeated measures.

OBJECTIVE: To examine the effect of the method of data display on physician investigators' decisions to stop hypothetical clinical trials for an unplanned statistical analysis. DESIGN: Prospective, mixed model design with variables between subjects and within subjects (repeated measures). SETTING: Comprehensive cancer centre. PARTICIPANTS: 34 physicians, stratified by academic rank, who were conducting clinical trials. INTERVENTIONS: PARTICIPANTS were shown tables, pie charts, bar graphs, and icon displays containing hypothetical data from a clinical trial and were asked to decide whether to continue the trial or stop for an unplanned statistical analysis. MAIN OUTCOME MEASURE: Percentage of accurate decisions with each type of display. RESULTS: Accuracy of decisions was affected by the type of data display and positive or negative framing of the data. More correct decisions were made with icon displays than with tables, pie charts, and bar graphs (82% v 68%, 56%, and 43%, respectively; P=0.03) and when data were negatively framed rather than positively framed in tables (93% v 47%; P=0.004). CONCLUSIONS: Clinical investigators' decisions can be affected by factors unrelated to the actual data. In the design of clinical trials information systems, careful consideration should be given to the method by which data are framed and displayed in order to reduce the impact of these extraneous factors.

Mississippi mud in the 1990s: risks and outcomes of vancomycin-associated toxicity in general oncology practice.

BACKGROUND: Discrepancies between the severity of toxicities reported in early clinical trials and recent clinical experience with vancomycin have led to confusion regarding the need for routine serum vancomycin level monitoring and discontinuation of vancomycin when toxicities occur. Therefore, the authors examined the incidence, outcomes, and predictive factors of vancomycin-associated toxicities in general oncology practice with the goal of developing clinically relevant prediction rules and guidelines. METHODS: All 742 consecutive cancer patients who received vancomycin at a comprehensive cancer center during a 3-month period were followed prospectively for the development and outcome of phlebitis, rash, ototoxicity, and nephrotoxicity. Logistic regression was used to derive a multiple variable model of the risk of nephrotoxicity. A clinical prediction rule, the Nephrotoxicity Risk Score, was developed from the risk model and validated prospectively. RESULTS: Phlebitis occurred in 3% of patients (95% confidence interval [95% CI], 2-4%), predominantly those with recently inserted central venous catheters. Rashes occurred in 11% of patients (95% CI, 9-13%); however, all but 4 patients also were receiving beta-lactam antibiotics. Clinical evidence of ototoxicity developed in 6% of patients (95% CI, 4-9%) who were receiving vancomycin plus other ototoxic agents and only 3% of patients (95% CI, 2-5%) not receiving other ototoxic agents (P = 0.08). Nephrotoxicity occurred in 17% of patients (95% CI, 15-20%). Logistic regression revealed that factors associated with an increased risk of nephrotoxicity included administration of other mild to moderate (P = 0.01) or severely nephrotoxic agents (P < 0.001) or an acute physiology and chronic health evaluation (APACHE) score > 40 (P = 0.002). Elevated serum vancomycin peak levels did not reliably predict subsequent nephrotoxicity. CONCLUSIONS: Vancomycin-associated toxicities usually are mild and self-limiting. Some patients are at a significantly higher risk of nephrotoxicity but the authors believe these individuals can be identified reliably with the Nephrotoxicity Risk Index using information available at vancomycin initiation. Further testing of the Nephrotoxicity Risk Index is ongoing.