"Killing, Determination of Fatality, Necropsy and Histopathology An equal number of rats of each sex were randomly selected from groups receiving 10 mg/kg ENU and killed by carbon dioxide inhalation at 30-day intervals between 171 and 294 days of age. Similarly, rats receiving 6.25 mg/kg ENU were to be killed at 30-day intervals between 199 and 325 days of age. Ten rats from each group of 60 (30 females and 30 males) were to be killed at each scheduled interval, but the numbers varied because of ENU-induced deaths and moribund killings and because of several technical errors, including the premature euthanasia of 610 ZOOK AND SIMMENS 30 6.25-mg/kg rats on days 171–174 instead of days 199–202 (Table 1). A total of 257 rats died or were killed humanely when they exhibited neurological signs or became moribund. A rat was considered to have a fatal tumor if it became moribund with clinical signs consistent with a particular tumor that was confirmed at necropsy. Rats that died without observable clinical signs had to have convincing necropsy evidence that a particular tumor caused death. All rats were necropsied, and the entire brain and spinal cord were removed, fixed in 10% formalin, processed for histology, and examined microscopically. All brain tissues were examined by the study and reviewing pathologists, who were blinded as to the exposure group, mode of death, and date of death or killing. Differences of opinion between the study and review pathologist occurred in less than 1% of tumors and were resolved by mutual agreement after further study. Other tissues similarly processed and examined microscopically from all rats were thyroids, pituitary, adrenals, liver, kidneys, lungs, spleen, heart, trigeminal nerves and any other tissues that appeared abnormal. Tissues from rats that died were never autolyzed to the degree that proper diagnosis was impossible. Spinal and cranial nerve tumors were classified as benign or malignant by virtue of their gross and microscopic features of anaplasia and invasiveness. Brains were sliced into transverse sections and were processed by taking histological sections 1 mm apart. Brain tumor volume was determined with the aid of computerized morphometry (39). The entire spinal cord was sectioned in transverse and longitudinally oriented portions and studied histologically. All tissues were stained with hematoxylin and eosin, and immunochemistry (GFAP, S-100, vimentin, Leu-7) was performed on 59 selected brain tumors as described previously (24, 39). The histological classification of tumors, measurement of volume, and determination of multiplicity and fatality and were performed as described previously (24, 39). The relative malignancy of brain tumors was determined by estimating the degree of anaplasia, invasiveness, vascular endothelial proliferation, necrosis, hemorrhage and displacement of surrounding brain structures. Each of these characteristics was graded on a scale of 0 to 3. The grade of each histological feature was multiplied by a factor (weight) consigned to each feature, and the result was totaled to give the malignancy score (0 to 100) for every tumor. The weight given to each histological characteristic was based on the relative degree to which each characteristic differentiated between fatal and nonfatal tumors, as determined in a previous study (39). These weights were modified here based on further data to the following values: anaplasia, 3.6; invasion of tumor borders, 3.7; invasion of ventricles, 2.8; vascular endothelial proliferation, 4.1; focal necrosis 5.0; hemorrhage, 7.0; and herniation, 7.0. Tumor characteristics of mitotic index and cellularity, used previously, were deleted since further study revealed a relatively poor correlation with fatality. Statistical Methods All data distributions were reviewed for outliers, out-of-range values and distributional form. Tests of differences between the three exposure groups on continuous outcomes were through analysis of covariance, controlling for potentially important determinants of outcome. Analysis of covariance P values were adjusted for multiple comparisons using the Tukey-Kramer procedure (49). The primary study hypothesis was that the pulsed RF-signal exposure group would have a higher observable brain tumor incidence rate (or equivalently a shorter latency) compared to the sham and cage control groups. Brain tumors were observed at various planned and unplanned times in this study and did not occur in all rats; therefore, time-to-event analysis methods that account for both the time until the tumor was observed and censoring were required. These methods, often known by the term ‘‘survival analysis’’, were used here to compare the study groups on relative incidence rates for development of a brain tumor. The age of the rat at the time of death was considered a censored time if no brain tumor was observed, regardless of whether the death was a preplanned killing or not. Kaplan-Meier curves were used to describe relative differences among study groups. Statistical tests of differences in incidence rate were conducted in two steps using the Cox proportional hazards regression model. The first step provided an omnibus test of any differences among the three groups using the Score Test. For the second step, potentially important determinants of tumor incidence were added as covariates to the model. The Cox proportional hazards model was used in the second step to estimate relative hazard rates and calculate P values and confidence intervals contrasting the groups while controlling for these covariates. Graphical methods were used to evaluate the proportional hazards assumption of these models. The above procedures were repeated for the other two tumor types commonly induced by ENU—spinal nerve and cranial nerve tumors. Comparisons of the exposure groups on the percentages of tumors found to be malignant were tested using the Pearson x2 test. The unit of analysis was the tumor rather than the rat. Note that the resulting P value is somewhat lower than an exact P value because some tumors come from the same rats, violating the independent observation assumption. Therefore, a more stringent a 5 0.01 was used for those tests rather than the conventional 0.05 level. All analyses were conducted using SAS/ STAT version 9.1.3 (49). The study sample size was planned so as to have good power to detect a small to moderate difference in brain tumor incidence between the RFsignal group and either control group. For example, the sample size of 360 rats in each group provides 92% power to detect a final 62% brain tumor rate in the RF-signal group in contrast to a hypothetical 50% rate in a control group, assuming a 5 0.05 (two-sided) and a two-sided logrank test applied to exponentially distributed survival times "

Sprague-Dawley rats, PRF

HAWC