One of the most striking characteristics of pediatric BL is the variant frequency of EBV association in different geographic regions . In tropical Africa it is almost always EBV-related. Conversely, in sporadic BL in developed countries, EBV association has been demonstrated in 15 to 30% of cases [5, 22]. The association of BL and EBV in developing countries is intermediate between the sporadic and endemic types [8, 23, 24]. In South America, a high association of EBV and BL was reported in the Northeast of Brazil (~80%) [25–27], and a lower association was observed in patients from Argentina and Chile [28–30]. In the present study, we detected a frequency of EBV association of 68% in 41 childhood NHL from Southeastern Brazil, which is higher than in developed countries. Thus, the use of EBV for identifying new therapy targets in poor-risk, EBV-positive lymphomas is of interest, leading to an effort to improve current EBV diagnosis.
EBER-RNA in situ hybridization is the standard for EBV diagnosis in tumor cells  while PCR procedures are used for EBV typing. Although the simplicity of PCR might favor its adoption as a first-line method for diagnosis, its high sensitivity may produce false positive results due to detection of EBV-positive memory cells and/or non-tumor, bystander lymphocytes. However, the frequency of EBV-positive memory cells in healthy seropositive individuals accounts for less than 1/50,000 in almost all estimates [3, 4, 31] while BL and DLBCL are mainly characterized by lymphoproliferation of monomorphic cells carrying a high viral load when infected by EBV . These observations prompted us to validate PCR assays for EBV diagnosis in these high-grade NHL, where strictly standardized PCR methods may have an important role. First, in a high complexity center offering cancer care to a large population, RISH is a second or third timeline method, considered only after histopathological and immunohistochemical diagnosis. The early definition of EBV status is important in some situations, for instance, to define the enrolment of a patient in a clinical protocol or in a viral load monitoring study, or to proceed with biological studies on fresh material. Second, the financial cost of RISH may be limiting in low-resource countries, making the effort to develop a rationale for EBV diagnosis, including PCR as a first-line, rapid approach followed by RISH for confirmation, significant.
We present a specific and reliable method for EBV-detection and typing in clinical samples. The choice of PET-DNA as PCR template aimed to make the results of both methods comparable, and to test the reliability of PCR in the most unfavorable technical conditions. Our comparisons of EBV detection by RISH and PCR showed that both methods provided highly concordant data (95%) with the same sensitivity (96%). The low EBV detection in reactive lymphoproliferations (4%) also points to the suitability of adopting PCR as a routine screening test, which can be instrumented based on single round PCR, decreasing the risks of potential cross-contamination.
Experiments of nested-PCR assays with Namalwa well-preserved DNA showed the highest sensitivity of our method. Even this sensitivity excluded the possibility of detecting EBV-bearing memory cells in clinical samples, reinforced by paired analyses of 9 BM and tumor samples, in which EBV was detected by PCR only in cases showing BM infiltration by EBV-positive tumor cells. It should be mentioned that the proposed diagnosis scheme depends on the sensitivity of our PCR method, because a more sensitive method might lead to a different conclusion.
In previous PCR typing studies in BL, concordant RISH/PCR results were reported in RISH-positive cases that were subsequently PCR-tested [23–30]. As no information was provided on PCR results in EBER-negative cases, an exceedingly high sensitivity of those EBV-specific PCR methods, making them unsuitable for diagnostic purposes, could not be assessed.
In this study, although only one viral gene was tested, viral DNA appeared to be more efficiently amplified than PET genomic DNA, raising the possibility that viral DNA might be more resistant to degradation, probably due to a differential effect of the tissue fixative on EBV DNA/protein complexes . However, if PCR is used as a screening test, amplification of cellular genes should be considered mandatory.
The molecular detection of EBV in lymphomas has not produced coherent results due to biological heterogeneity and to methodological differences [33–35]. The diagnostic criteria recommended by the IARC-WHO  are appropriate for entities like HD, where the heterogeneity of clinical subtypes and cell types justify this conservative approach . Molecular EBV detection should also be cautiously considered in peripheral T-cell NHL and anaplastic large-cell lymphoma, since they comprise heterogeneous entities with uncertain clonality status [36, 37].