Many treatment strategies have been proposed for pyogenic liver abscesses; however, the indications for liver resection for treatment have not been studied in a systematic manner. The purpose of our study was to evaluate the role of surgical treatment in pyogenic abscesses and to determine an optimal treatment algorithm. We retrospectively reviewed the medical records of all patients who had a pyogenic liver abscess at Rhode Island Hospital between 1995 and 2002. Abscesses and treatment strategies were classified into three groups each. The abscess groups included Abscess Type I (small <3 cm), Abscess Type II (large >3 cm, unilocular), and Abscess Type III (large >3 cm, complex multilocular). The treatment strategy groups included Treatment Group A (antibiotics alone), Treatment Group B (percutaneous drainage plus antibiotics), and Treatment Group C (primary surgical therapy). Descriptive statistics were calculated and chi2 used for comparison with a P < 0.05 considered significant. Our study consisted of 107 patients with pyogenic liver abscess. The success rate for small abscesses treated with antibiotics was 100 per cent. The success rate with antibiotics and percutaneous drainage for large, unilocular abscesses was 83 per cent and for large, multiloculated abscesses was 33 per cent. None of the 27 patients who had surgical therapy for large, multiloculated abscesses had recurrences. Surgical treatment for large (>3 cm), multiloculated abscesses had a significantly higher success rate than percutaneous drainage plus antibiotic therapy (33% versus 100%, P < or = 0.01). The mortality rate for the percutaneous drainage plus antibiotic group was not significantly different from the primary surgical group (4.2% versus 7.4%, P = 0.40). We propose a treatment algorithm with small abscesses being treated with antibiotics alone; large, uniloculated abscess with percutaneous drainage plus antibiotics; and large, multiloculated abscessed treated with surgical therapy.