Lytic units reconsidered: pitfalls in calculation and usage.

51Chromium release-derived cytotoxicity data yield curvilinear plots when the x axis displays the effector:target ratio and the y axis displays the percentage of cytotoxicity. To facilitate data analysis, several biomathematical models (simple linear regression, exponential fit, and Von Krogh) have been used to express these cytotoxicity curves as a single numerical value, termed the lytic unit. Other than using raw cytotoxicity data, the lytic unit has been the most common method of data presentation in human and animal tumor immune studies involving natural killer cells, lymphokine-activated killer cells, and cytotoxic T cells. Unfortunately, the models for determining lytic unit values incorporate assumptions and methods of calculation that can result in inaccurate model-predicted cytotoxicity in comparison with the actual observed cytotoxicity data. Even when the model is accurate in predicting cytotoxicity values (i.e., the nonlinear regression-calculated three-parameter Von Krogh model), comparisons between donors of minimally different or highly different cytotoxicity are still fraught with potential error due to statistically verifiable violations of assumptions of parallelism. Although more cumbersome, donor cytotoxicity comparisons using a range of effector:target ratios are not subject to the above problems. Researchers may therefore want to reconsider the use of lytic units when evaluating and reporting cytotoxicity data.