Materials and Methods

Cattle—A convenience specimen of 25 adult cattle that were evaluated for systemic illness and had enlarged peripheral lymph nodes was used in the study. All 25 cattle were necropsied. Owners consented to enrollment because the cattle had a terminal disease, and this protocol was approved by the Oklahoma State University Animal Care Use Committee. A peripheral lymph node was deemed enlarged on the basis of the experience of the examining clinician. Three clinicians participated in the study and selected cattle for enrollment.

Antemortem testing—Clinical evaluation of the cattle consisted of a thorough physical examination including transrectal palpation. Other diagnostic tests were conducted on the basis of the problem-oriented approach to case diagnosis and management and included hematologic and serum biochemical analyses as well as abdominocentesis and pericardiocentesis. All tests were not performed on all cattle because of varying clinical findings or economic constraints of the owners.

Biopsy and FNA—In each animal, 1 enlarged lymph node was selected to be sampled. In cattle in which 1 lymph node was larger than others, it was selected; if all were equally enlarged, selection was based on the ease of accessing the node in the cattle chute. The particular lymph node to be sampled was documented in the record (prescapular or prefemoral). The lymph node was measured dorsal to ventral at its highest and lowest points, respectively, and cranial to caudal with a flexible metric tape measure. Cattle were restrained in a chute, and the skin over the lymph node was clipped and aseptically prepared. A small bleb (2 mL) of 2% lidocaine was placed in the skin over the lymph node in the area of the CNB and FNA. A No. 15 scalpel blade was used to make a stab incision through the skin at the location of the lidocaine bleb. Specimens were collected via FNA first, followed by collection of the CNB specimens.

For the FNA, a 20-gauge, 4-cm needle attached to a 12-mL syringe was placed through the stab incision into the lymph node. Approximately 4 aspirates were performed by drawing back on the syringe plunger. Two aspirates were performed in a horizontal plane to the lymph node, and 2 aspirates were performed in a plane vertical to the lymph node. If blood appeared in the hub of the needle, the needle was discarded and another was used. After aspiration, the needle was withdrawn from the lymph node without negative pressure. Air was drawn into the syringe through the needle, and the syringe contents were quickly forced out onto a microscope slide. Once applied, the specimen was distributed across the slide by use of a squash preparation technique,⁷ air dried, and prepared for cytologic examination with Wright stain.

For the biopsy specimens, a 14-gauge, 12-cm CNB instrument^a was placed through the same stab incision into the enlarged lymph node while the lymph node was stabilized with the opposite hand. Biopsies were performed in this manner in 2 planes (horizontal and vertical to the lymph node) to obtain at least 2 specimens of tissue. Biopsy specimens were fixed in neutral-buffered 10% formalin and routinely processed for histologic examination with H&E stain. All slides were labeled with a hospital case number and stored for interpretation after final necropsy results were obtained. At necropsy, pathologists performed a routine thorough examination and obtained specimens from various organs and peripheral lymph nodes as deemed necessary to determine a definitive diagnosis.

Analysis of specimens—After all antemortem and postmortem data had been collected, the CNB specimens were renumbered from 1 to 25 and submitted to 3 anatomic pathologists for interpretation. The pathologists were informed that the specimens were from clinically enlarged lymph nodes; however, they were unaware of other clinical and diagnostic findings and definitive postmortem results. Each specimen was interpreted as reactive, neoplastic, or nondiagnostic. The pathologists used a scale from 1 to 4 to rate their confidence in interpretation, with 1 being least confident and 4 being most confident.

The FNA slides were also renumbered from 1 to 25 and submitted to 3 clinical pathologists (different individuals than the anatomic pathologists) for evaluation. The clinical pathologists evaluated the specimens under the same conditions and with the same criteria as the anatomic pathologists.

The corresponding necropsy reports were obtained and analyzed in regards to final results and diagnosis. In particular, information regarding the lymph nodes examined and observations in the nodes was obtained. Findings from antemortem specimens were analyzed and compared with postmortem results.

Statistical analysis—Diagnostic test properties for CNB and FNA were based on the postmortem diagnoses as the gold standard. Values for sensitivity, specificity, positive predictive value, negative predictive value, overall test accuracy, and corresponding 95% confidence intervals were calculated.^b Specimen prevalence values were used for predictive value calculations. Likelihood ratios for positive and negative test results were calculated.^c For evaluations with zero counts in cells of the 2 × 2 tables, a value of 0.05 was used to replace the zero so that the likelihood ratio confidence interval could be calculated without encountering a division-by-zero error in the formula. Agreement beyond chance among the CNB results from the pathologists and among the FNA results from clinical pathologists was assessed by use of the kappa statistic.^d The Somers d statistic was used as a measure of ordinal association to evaluate the effect of confidence level provided by the pathologist or clinical pathologist on the accuracy of CNB and FNA results, respectively, compared with the postmortem findings.^d This evaluation was performed for all cases and for a subset of cases limited to EBL diagnosis only, and P values were based on exact significance calculations. Spearman correlation coefficients with 2-tailed P values were calculated to determine any significant associations between test result accuracy for CNB and FNA and postmortem examination of the same or possibly a different peripheral lymph node.^d For all statistical tests, the level of significance was P < 0.05.

Results

Twenty-five cattle of both sexes were enrolled in the study. Age ranged from 6 months to 11 years, and various breeds were represented, including both Bos taurus (beef and dairy) and Bos indicus.

During the physical examination, all accessible peripheral lymph nodes were palpated. Lymph nodes deemed to be enlarged were right and left prescapular and prefemoral nodes. Prescapular lymph node measurements ranged from 5.5 to 17 cm in length, 4 to 11 cm in width, and 3 to 6.2 cm in depth. Prefemoral lymph node measurements ranged from 7 to 15 cm in length, 4 to 5 cm in width, and 3 to 7 cm in depth.

A prescapular lymph node was selected for sampling in 13 cases, and a prefemoral lymph node was selected in 12. All 3 pathologists interpreted the same 3 CNB specimens as nondiagnostic because of the condition of the tissue or lack of lymph node tissue. Each pathologist also interpreted from 1 to 3 other specimens as nondiagnostic. Therefore, CNB results were analyzed from 19 to 21 specimens, depending on the individual pathologist. For FNA, all 3 clinical pathologists interpreted 2 specimens as nondiagnostic and aspirates were not obtained in another instance; therefore, results from 22 specimens were analyzed.

Nineteen true EBL cases and 6 non-EBL cases were diagnosed via necropsy. In the case of OSCC, findings were considered inflammatory because of severe necrosis, suppuration, and invasion of tumor into deep tissues in the ocular region with no histologic evidence of OSCC in the lymph node.

Point estimates and 95% confidence intervals for variables used to evaluate performance of the CNB and FNA were determined (Table 1). On the basis of the number of specimens determined to be diagnostic by each individual pathologist or clinical pathologist, specimen prevalence values for EBL were from 75% to 79% for CNB and 77% for FNA. These prevalence values were used to calculate results for positive and negative predictive values. For CNB, sensitivity ranged from 38% to 67% and specificity ranged from 80% to 25%. In comparison, for FNA, sensitivity ranged from 41% to 53% and specificity was 100% because no specimen results were misclassified as positive (neoplastic) from cattle that did not have EBL. The predictive values for positive test results were from 77% to 89% for CNB and 100% for FNA. Predictive values for negative results were low for both CNB and FNA, ranging from 17% to 38%, which reflected from the finding that results from 33% to 63% of specimens from cattle with EBL were classified as negative. For FNA, likelihood ratios for positive results were > 41, which provided large, conclusive changes from pretest to post-test probabilities for EBL. By comparison, CNB likelihood ratios for positive results were < 3, which provided small to negligible changes in posttest probabilities relative to the pretest situation. In fact, for one of the pathologists, the likelihood ratio for positive results was < 1, which provided no advantage over the pretest status. For negative test results, both CNB and FNA yielded small to negligible changes in the likelihood ratios, with the values ranging from 0.48 to 1.33.

Evaluations of all possible pairwise comparisons among pathologists (CNB results) and clinical pathologists (FNA results) to evaluate agreement beyond chance provided significant (P = 0.02) results for all comparisons. There was fair to moderate agreement among pathologists for CNB results, as indicated by kappa statistic values that ranged from 0.37 to 0.43. For clinical pathologists, there was good agreement of results, as indicated by kappa statistic values ≥ 0.85.

Evaluation of the level of confidence assigned by the diagnostician for CNB and FNA regarding accuracy of diagnosis revealed that there was no significant association whether all cases were considered or evaluations were limited to those cattle that had EBL. The closest any evaluations came to being significant for the effect of confidence on outcome was for 1 pathologist with P values of 0.12 and 0.08 for all cases and a subset of EBL cases, respectively. Evaluations for the other pathologists and all of the clinical pathologists provided P values > 0.33.

Necropsy reports revealed that in 17 cattle, the pathologist specifically examined and described the findings in the same peripheral lymph node that was used for obtaining specimens antemortem. In the remaining 8 cattle, it was uncertain whether the same lymph node used for antemortem specimen acquisition was examined histologically after necropsy. Spearman correlation coefficients indicated that there were no significant associations between test result accuracy for CNB and FNA and postmortem examination of the same or possibly a different peripheral lymph node (P = 0.37). Of the cattle in which EBL was the final diagnosis, 13 were cattle in which histologic examination was performed on the same lymph node used for antemortem specimen acquisition. In 6 cattle, postmortem diagnosis of EBL could have been derived from examination of different lymph nodes than those used antemortem. A comparison of CNB and FNA results for each diagnostician for all cattle and a subset of cattle for which the same lymph node was used antemortem and evaluated at necropsy revealed no significant differences in test accuracy. Depending on the particular pathologist, this comparison of CNB results provided a full spectrum of results that varied from small gains and small losses in sensitivity and specificity as well as a small gain in sensitivity combined with a small loss in specificity. The percentage differences ranged from 2% to 8% with a mean of 5.2%. Changes in predictive values had a pattern similar to changes in sensitivity and specificity. For FNA results, improvements in sensitivity ranged from 2% to 4% among the clinical pathologists with no difference in specificity for these comparisons because of the 100% overall specificity. Gains in predictive values for negative results for the subset of cases ranged from 6% to 7%. None of these differences in FNA results were significant.

Discussion

Cattle were selected in this study on the basis of clinical findings of terminal or untreatable disease with enlargement of peripheral lymph nodes. Because of variations in experience of the clinician, body condition of the animal, and clinical judgment, the authors were aware of the subjectivity of the selection process with regards to lymph node enlargement. The intent was to select a mixture of cases, both neoplastic and inflammatory, to model a clinical scenario in which antemortem diagnosis of EBL was difficult. More cattle with EBL were included in the study than those with inflammatory conditions. In cattle in which one of the peripheral lymph nodes was larger than the others, it was selected for sampling. In some cattle, that lymph node appeared to be draining an adjacent injection site lesion or, in 1 animal with OSCC, a suppurative, necrotic process secondary to neoplasia. These nodes were selected to truly test the ability of a pathologist to differentiate neoplasia from inflammation. The submandibular lymph nodes were not selected for sampling in any of the cattle because of the presence of many vital structures in the area that would have made CNB and possibly FNA of these nodes more difficult and potentially harmful to the animal.

Bovine leukemia virus status was not determined in any of the cattle enrolled in the study. This was omitted because extensive study of its usefulness as an antemortem tool to diagnose enzootic lymphosarcoma has been performed. However, adding results of a BLV gp51 antigen test to the results of other ancillary tests and lymph node biopsy or FNA would certainly increase the confidence in an antemortem diagnosis.

Three of the CNB specimens and 2 of the FNA specimens were nondiagnostic according to all 3 anatomic and clinical pathologists. This was most often attributable to lack of lymph node tissue in the specimen; muscle and connective tissue were most often observed in these specimens. This reflects the inherent difficulty in obtaining a percutaneous CNB or FNA specimen from a bovine lymph node. The authors found that the thick skin and the capsule surrounding the lymph node created difficulties in obtaining a sufficient specimen in modestly enlarged nodes. Use of ultrasound guidance may have increased the probability of obtaining lymph node tissue.

Three anatomic and 3 clinical pathologists independently examined the specimens and were, at the time of the study, providing diagnostic services to the university teaching hospital and the Oklahoma Animal Disease Diagnostic Laboratory. It was thought that because of the differences in interpretation of lymph node tissue specimens among pathologists, examination by more pathologists would improve the confidence of a clinician that the true diagnosis had been reached. Unfortunately, from 30% to 40% of EBL cases were not correctly diagnosed by any of the 3 evaluations of CNB or FNA specimens, which indicates the difficulty of making a correct diagnosis of a systemic disease on the basis of a limited specimen of tissue or cells. In addition, differences in pathologist experience may have explained the variations in confidence levels. In this study, the confidence of the diagnostician was not correlated with accuracy of the antemortem diagnosis.

The fact that only a small core specimen of lymph node was obtained for examination could explain differences between the histologic interpretation made antemortem and the final necropsy results. It is likely that in some cattle, a specimen was removed from a portion of the lymph node that was not affected by neoplasia and the pathologic interpretation was indeed correct for that portion examined.

One debatable aspect of this study was the inclusion in the statistical analysis of the 8 cattle in which it was not certain that the lymph node examined postmortem was the same node used for acquisition of specimens antemortem. It could be argued that exclusion of these cattle would provide a more precise evaluation of the true ability of FNA and CNB to provide the proper diagnosis when the lymph node selected antemortem was known to contain or not contain neoplasia as confirmed via necropsy. However, we believe that inclusion of these cattle strengthened the clinical applicability of the study because, although pathologists are not restricted in the tissues selected to arrive at a definitive diagnosis of EBL, clinicians are limited to a small set of available tissues to examine, namely peripheral lymph nodes.

Results of this study indicated that CNB and FNA may be better suited as confirmatory tests to support a diagnosis of EBL because of their relatively high specificity, compared with sensitivity, especially for FNA. Likewise, these tests would perform poorly as screening tests because of their inherently low sensitivity.

In most instances of comparison of accuracy and performance, FNA is superior to CNB, although the small number of cattle evaluated in this study likely precluded detection of significant differences. However, it is interesting that in a similar study⁸ that compared FNA to CNB for evaluation of palpable mammary lesions in humans, there was a similar advantage for FNA. The present study provided an important finding in that there were no false-positive results for FNA, which resulted in 100% specificity. This was especially noteworthy considering that all 3 clinical pathologists were consistent for this result.

The hypothesis that CNB specimens could provide sufficient tissue for diagnosis of neoplasia was supported; however, the hypothesis that FNA would not be a useful aid in the antemortem diagnosis of EBL was not supported. Instead, results provided evidence that FNA may perform as well or perhaps better than CNB for diagnosis of EBL in cattle.