Most hand-held medical refractometers have internal scales that limit protein measurement to results >/=2.5 g/dL. Tables for conversion of refraction (r) to protein concentration for values as low as 0.1 g/dL were published in the 1960s, but their accuracy for use on body fluids has not been established. The purpose of this study was to assess the reliability of body cavity fluid protein determination by refractometry. We compared the protein concentration of 25 body cavity fluids as determined by 2 Goldberg type hand-held refractometers with results obtained by the biuret method. Published charts converting refraction (r) to protein concentration were used to determine protein concentration in samples with protein <2.5 g/dL. Higher protein values were read directly from the instruments. The range of comparison was limited to >/=0.6 g/dL, the lowest concentration of the biuret method's standard curve. Twenty-one peritoneal fluid, 2 pleural fluid and 2 pericardial fluid samples from 16 horses, 5 cattle, 3 dogs, 2 llamas and 1 cat were tested. The results obtained by the two refractometers were closely and linearly related to biuret results (P<.001), with slopes by linear regression analysis close to 1, and correlation coefficients >0.977. Based on this study, the range for quantification of body cavity fluid protein concentration by refractometry can be extended below 2.5 g/dL, allowing for quantitative assessment of most clinical samples.