The gamma evaluation method is widely used for accurate quantitative comparison of measured dose distribution and treatment plan verification in intensity-modulated radiotherapy. In the case of field-in-field (FIF) plans, the standard gamma criterion fails to distinguish clinically significant differences in planned and actual dose distributions. The proposed modified gamma evaluation algorithm takes into account only high-dose-gradient regions in calculating the value of the gamma criterion. The aim of this study is to test the new algorithm in FIF plans, evaluating the test's sensitivity to dose, segment size and shape, and compare it with a commercially available gamma algorithm. For this study, several clinical cases were modeled using MATLAB® software. For segment shape and size tests a 20 × 20 cm base field was modeled. Randomly shaped rectangular segment fields with sizes from 5% to 50% of the base field area with 5% increments were added to the base field. Dose distribution was calculated using MATLAB® for use as a starting point to make distorted plans with segments displaced to simulate patient movement or positioning errors during irradiation. For sensitivity tests a 20 × 20 cm base field and rectangular segment with an area of 5% of the base field was added at the center. For the reference, 120 dose units were prescribed, 100 for the base field and 20 for the segment. Dose modulation was done by altering the segment dose in steps of two dose units per measurement. Segment dose-modulated fields were compared with the reference using the modified gamma criterion. Pixel acceptance criteria were based on dose difference DM = 3% and distance-to-agreement (DTA) dM = 3 mm. Only pixels in those regions where the dose gradient exceeded 5% were used; and 95% of all pixels in the region should be within this criterion for the treatment plan to be accepted. Comparison with commercially available gamma evaluation software demonstrated that the standard criterion is not able to detect clinically significant shifts of 5 mm in the FIF plans, but the new algorithm detected all shifts before they exceeded 3 mm. In turn, the commercial gamma criterion showed a higher dose sensitivity: dose alterations larger than two dose units show a steep drop of 'passed' pixel number below the 95% threshold level.