We have reviewed our experience with 43 cases of bacterial spinal epidural abscess, as well as previously reported series of cases. We found a striking male predominance of the disease, accounting for 86% of cases. Most patients had some underlying conditions that predisposed to infection, a prior infection at a distant site, or an abnormality or trauma to the spine. Presenting symptoms included backache (72%), radicular pain (47%), weakness of an extremity (35%), sensory deficit (23%), bladder or bowel dysfunction (30%), and frank paralysis (21%). Patients cared for in public hospitals tended to seek medical attention in later stages of the disease than patients admitted to private hospitals. Spinal epidural abscess was the suspected diagnosis in only 40% of the cases; the remainder of the time various other infections, tumors, neurologic diseases, or degenerative conditions were considered. Patients in whom the diagnosis of spinal epidural abscess was not initially entertained on admission suffered delays in diagnosis and experienced neurologic deterioration. Staphylococcus aureus was the predominant pathogen (65%) and was associated with positive blood cultures in nearly every case; aerobic or facultative gram-negative bacilli were next most common. Coagulase-negative staphylococci caused infection only in patients who had previous spinal instrumentation. Although analysis of CSF was abnormal in the majority of cases, abnormalities were nonspecific, Gram stain was always negative and culture was rarely diagnostic. Abscesses extended over an average of 4 vertebrae, and the majority were located in the lumbar region followed by thoracic and cervical regions. Unlike previous series, we noted an equal frequency of anterior and posterior epidural abscesses; although differences were not statistically significant, posterior abscesses tended to be more extensive but less commonly associated with radiographic abnormalities of osteomyelitis. Myelography revealed an abnormality in every case in which it was done. Computerized tomographic scanning after intrathecal injection of contrast material always provided additional useful information. Even though magnetic resonance imaging was diagnostic in only 4 of 5 cases (80%) in our series, this test is noninvasive and clearly delineates the location and nature of spinal lesion. It should, therefore, probably replace myelography as an initial definitive study in patients suspected of having spinal infection. Plain roentgenograms and nuclear scans contributed little useful information that was not already available from other radiographic procedures. Surgical drainage together with antibiotics was the treatment of choice; 35 of our 43 patients underwent operative intervention. The preoperative status clearly predicted the final neurologic outcome.(ABSTRACT TRUNCATED AT 400 WORDS)