Estimating hypoxic status in human tumors: A simulation using Eppendorf oxygen probe data in cervical cancer patients

Purpose: To define the minimal number of pO(2) measurements, with 90 % sensitivity and 90 % specificity, needed to categorize cervical tumors as either hypoxic or oxic. Methods and Materials: Using Eppendorf oxygen probe data from our ongoing prospective trial, we simulated the measurement of tumor oxygenation with a smaller number of data points in 135 patients with cervical cancer. The hypoxic proportion, defined as the percentage of pO(2) values <5 mm Hg (HP5), was calculated for each tumor. Hypoxic tumors were defined as those with a median HP5 > 50 %, and tumors with normal oxygen levels as those with a median HP5 less than or equal to50%. A small number of pO(2) measurements were randomly selected from the Eppendorf measurements in each tumor, or per Eppendorf track, and used to define the tumor as hypoxic or oxic. The sensitivity and specificity were calculated, considering the classification as given by the complete set of Eppendorf measurements as the reference standard. Results: The probability of falsely classifying the tumor decreased as the selected number of pO(2) measurements per tumor increased, and at 16 measurements was approximately 10%. Adding additional measurements per tumor beyond 24 improved the ability to classify the tumor accurately only slightly. The probability of falsely classifying the tumor decreased as the pO(2) measurements per track increased. At five measurements per track, the probability of falsely classifying the tumor was approximately 9%. Conclusion: Approximately 20 measurements per tumor, or five measurements per track, using the Eppendorf pO(2) histograph, are sufficient to categorize cervical tumors as hypoxic or oxic. The results of this study will serve as a guide for research clinicians in the use of this and other systems in the assessment of tumor oxygenation in humans. (C) 2003 Elsevier Science Inc.