The morphological consequences of delayed posttraumatic brain hyperthermia (39 degrees C) after fluid percussion brain injury were assessed in rats. Sprague-Dawley rats anesthetized with 4% halothane and maintained on a 70:30 mixture of nitrous oxide:oxygen and 0.5% halothane underwent moderate (1.5-2.0 atm) traumatic brain injury with the injury screw positioned parasagittally over the right parieto-occipital cortex. At 24 hours after traumatic brain injury, the rats were reanesthetized and randomized into two groups in which either a 3-hour period of brain normothermia (36.5 degrees C, n = 18) or hyperthermia (39 degrees C, n = 18) was maintained. Sham-operated controls (n = 10) underwent all surgical and temperature-monitoring procedures. After the 3-hour monitoring period, the rats were allowed to survive for 3 days for light microscopic analysis or were injected with the protein tracer horseradish peroxidase and were perfusion-fixed 15 minutes later for light and electron microscopic analysis. At 4 days after traumatic brain injury, delayed posttraumatic hyperthermia (n = 12) significantly increased mortality (47%) and contusion volume (1.7 +/- 0.69 mm3, mean +/- standard error of the mean), compared to normothermia (n = 12) (18% mortality and 0.13 +/- 0.21 mm3 contusion volume) (P < 0.01, analysis of variance). At 15 minutes after the 3-hour hyperthermic period, the area of hemorrhage and horseradish peroxidase extravasation overlying the lateral external capsule was significantly increased (2.52 +/- 0.71 mm2, mean +/- standard error of the mean, versus 0.43 +/- 0.16 mm2) (P < 0.01), compared to normothermic rats. Examination of toluidine blue-stained plastic sections demonstrated a higher frequency of abnormally swollen myelinated axons per high microscopic field with hyperthermia. For example, numbers of swollen axons within the sixth layer of the right somatosensory cortex, corpus callosum, and internal capsule were 7.3 +/- 1.3, 4.2 +/- 1.4, and 3.0 +/- 1.2 axons (mean +/- standard error of the mean) with normothermia, respectively, compared with 24.7 +/- 12.1, 33.1 +/- 4.2, and 27.3 +/- 3.1 axons with hyperthermia, respectively (P < 0.01). An ultrastructural examination of the swollen axons demonstrated a severely thinned myelin sheath containing axoplasm devoid of cytoskeletal components. These experimental results indicate that posttraumatic brain hyperthermia might increase morbidity and mortality in patients with head injury by aggravating axonal and microvascular damage.