Determinants of diagnostic accuracy in pulmonary scintigraphy for pulmonary capillary protein leak associated with adult respiratory distress syndrome (ARDS): a technical note

Abstract

Radionuclide assessment of pulmonary capillary protein leak using [99mTc] human serum albumin (99mTc-HSA) was first reported from our laboratory. In this study we investigated the impact of 1) sampling time post tracer injection, and 2) lung region assignment, on diagnostic accuracy between 2 groups (control n = 20 and ARDS n = 20). Each patient received 370 MBq 99mTc-HSA i.v. and was imaged for 45 min. The slope index (SI) [change in lung: heart activity ratio/min] was calculated from 11 computer assigned lung regions for intervals of 5-15 (early [E]) and 15-45 (late [L]) min. The diagnostic accuracy of E vs L SI calculations for the 11 regions was evaluated by stepwise logistic regression. E SI data and L SI data from the lower 1/3 of the lung did not achieve significance for inclusion in the discriminant model (P less than 0.05). In the nine remaining regions L SI was significant. Optimal discrimination was achieved from L SI data obtained from a region confined to the lateral half of the mid 3rd of the lung field (sensitivity 81%, specificity 85%, accuracy 83%). The results confirm that: 1) a late (15-45 min) sampling period and 2) proper region assignment are necessary to maximize accuracy of this technique.