Kasner S E, Kimmel S E
Department of Neurology, University of Pennsylvania School of Medicine, Philadelphia, PA 19104, USA.
Cerebrovasc Dis. 2000 Jan-Feb;10(1):18-24. doi: 10.1159/000016020.
The clinical diagnosis of stroke subtype is often inaccurate during the first 24 h after stroke onset. Several candidate diagnostic tests might be useful for early determination of subtype, but there is no consensus on what level of accuracy is required to be useful in the selection of patients for subtype-specific studies or therapies.
A decision analysis was developed to explore the treatment options and estimate the requisite threshold of diagnostic accuracy. Four management strategies were considered: treat all (TA), treat based on new test (TBNT), treat based on clinical diagnosis (TBCD) and treat none (TN). Sensitivity analyses were performed over a wide range of the assumptions in the model.
The preferred treatment strategy was dependent on the probability and severity of adverse effects and the positive predictive value (PPV) of the proposed diagnostic test. For a potential therapy with infrequent (7.5%) but severe side effects, TN was preferable, but TBNT dominated if the PPV of the new test was at least 81%. For a therapy with frequent (25%) but mild adverse effects, TBNT was preferable. TA was favored for a therapy with infrequent and mild side effects and TN for a therapy with severe and frequent adverse events. TBCD was never the preferred option unless the PPV of the new test was less accurate than clinical diagnosis alone.
Clinical diagnosis of stroke subtype is insufficient for patient selection, but a new diagnostic test with PPV >/=81% may be useful for early subtype diagnosis and patient selection for stroke subtype-specific clinical trials.
