Although liquid-based cytology is the procedure most often used for cervical cancer screening, it has limitations related to subjectivity and relative insensitivity. The high-risk HC2 HPV DNA test is sometimes performed because of its advantages of high sensitivity and NPV. However, most women infected with HPV will eliminate the virus within 1-2 years, and only a very small percentage of them will progress to high-grade diseases. The HPV DNA test is therefore a diagnostic method with high sensitivity but low specificity, particularly for younger, sexually active women [19, 20]. It is now accepted that the integration of high-risk HPV into the host genome is one of the major contributing factors to cervical carcinogenesis, and this phenomenon can not only be observed in high-grade lesions but also in a proportion of low-grade lesions [21, 22]. The HPV DNA test is incapable of distinguishing HPV physical status (episomal vs. integrated) and is therefore not effective for identifying which patients with HPV infections are likely to have a CIN2+ lesion. According to the WHO guideline  and the 2006 consensus guidelines of the American Society for Colposcopy and Cervical Pathology (ASCCP) , the recommendation for CIN1 cases is to undergo follow-up examinations at defined intervals, whereas the recommendation for CIN2/3 cases is to undergo immediate treatment for the prevention of progression to carcinoma. In view of the limitations of the currently used screening methods, more accurate and reliable predictive biomarkers are needed to complement morphologically based differential diagnostic methods. For the differential diagnosis of low- vs. high-grade lesions, the change of the biomarker should ideally be an early event in cervical carcinogenesis that occurs in precancerous lesions.
Amplification of oncogenes is commonly observed in cervical precancerous lesion. It is a fairly early event in cervical carcinogenesis. Researchers have applied FISH probes to TERC and C-MYC, the two most frequently observed amplified oncogenes in cervical precancerous lesions as detected by CGH studies, for cytological specimen analysis. Heselmeyer et al. (2003)  applied a FISH probe set to cervical cytological specimens and found that the TERC gain cell counts and the maximum TERC copies was correlated with the severity of cervical lesions. After a follow-up of 1 to 3 years, the percentage of TERC amplified cases increased from 52% to 96% . Sokolova et al. (2007)  applied a TERC-MYC-HPV probe-mix to Thinprep slides and found that LSIL/HSIL cytological specimens with underlying CIN2/3 showed positive FISH test results in more than 80% cases at a cut-off value of 4 or more double-positive cells (HPV and TERC/MYC aberrations). Andersson et al. (2009)  applied a TERC-MYC-HPV probe-mix to Thinprep slides by a similar procedure but using different enumeration methods and cut-off values. All cells on the slides were enumerated, including HPV-positive and HPV-negative cells. This procedure increased the sensitivity by including the HPV-negative cells with oncogene amplification. The cut-off value of 9 cells with more than two TERC copies in a whole slide scan excluded the effect of HPV and MYC, and the cut-off value of the TERC test appeared to be higher than the cut-off value of Sokolova et al., thereby increasing the specificity of the FISH test.
The procedures for specimen processing and signal enumeration in our study differed somewhat from those of Sokolova et al. (2007) and Andersson et al. (2009). In our study, the slides for enumeration were prepared with cell suspension drops from residual Thinprep PreservCyt (Cytyc) cytological specimens. The cells were pre-treated with collagen B and deionized water and were thereby dispersed and enlarged. Because the cells were evenly mixed, the possible selection bias of enumeration vision was limited. The TERC probe and C-MYC probe were applied separately on two slides, in contrast to the two studies above. Regarding the comparison of enumeration methods, Sokolova et al. (2007) analyzed the entire surface area of each slide in most cases. However, when there was a large number of HPV-positive cells per slide, the first 100 HPV-positive cells were analyzed and the number of HPV-positive cells on a whole slide was extrapolated from the percentage of surface area occupied by the cells. Andersson et al. (2009) enumerated the signals by screening and counting the entire slide visually; an average of 2320 nuclei per slide (range: 232 to 4996 nuclei) were counted. Enumerating cells on a whole slide might be time-consuming and tedious. To achieve rapid and accurate screening for clinical usage, we enumerated only the first 100 cells on each slide and then screened the whole slide. The time cost for signal enumeration of a slide is determined by the number of cells on a whole slide and the complexity of hybridization patterns. It usually takes 30 to 60 min to evaluate both the TERC and C-MYC signals for one patient.
The cut-off value we used for CIN2+ diagnosis is 5% or more aberrant TERC cells, which is higher than the cut-off values used in the two studies above and is also higher than the cut-off value of ≥ 2.5% cells with more than 2 TERC signals used by Heselmeyer et al. (2003) . However, the cut-off value we used is consistent with that used in a multicenter study in China, and the mean TERC test cut-off value determined for all of the participating centers was 6.4 ± 2.3% . Regional and ethnic differences were not ruled out for cut-off value differences. Using this TERC test cut-off value, our study showed specificity values similar to those of Andersson et al. (89.6% and 83.9%, respectively) and higher sensitivity (90.0% and 78.7%, respectively) for CIN2+ diagnosis. As possible cut-off value choices for the C-MYC test, the values of ≥ 3% and ≥ 5% C-MYC gain cells showed similar AUC (0.8) and DFI (0.3) values. We chose ≥ 3% C-MYC gain cells as the cut-off value because of its higher sensitivity. Using this cut-off value, the C-MYC test showed a sensitivity of 80.0% and a specificity of 77.7%. The sensitivity and specificity of C-MYC would be similar to those of Andersson et al. if one used a cut-off value of ≥ 5% aberrant C-MYC cells (66.0% vs. 66.0%, and 94.8% vs. 87.1%, respectively).
High-level amplifications have certain indication roles for advanced-grade precancerous lesions. Compared with other types of solid tumors, cervical cancer has a relatively low-level of amplification, usually at 3 to 6 copies, and a minority of the nuclei have a GCN > 20. Tu et al.  found primarily TERC amplification patterns of 3 to 4 copies in the normal/CIN1 group, whereas in the CIN2+ group, the percentage of amplification patterns of 5 copies and 6 copies were 10.2% and 54.6%, respectively. In our study, we detected a SCC case with a TERC GCN of 22 and a CIN2 case with a C-MYC GCN of 12. We found that the percentage of abnormal nuclei increased with the severity of disease for both TERC and C-MYC (p < 0.05). The CIN2+ group showed more high-level TERC amplifications (GCN ≥ 7) than did the normal/CIN1 group. However, the C-MYC amplification patterns are similar between the normal/CIN1 and CIN2+ lesions. Although GCN is an indicator for CIN2+ diagnosis, the combined sensitivity and specificity of GCN is lower than that of aberrant cell percentage for CIN2+ diagnosis.
TERC amplification patterns become more diverse as histological grades increases. The formation of isochromosome 3q is frequently observed in cervical carcinogenesis. In CGH studies, gain of 3q and loss of 3p were usually observed simultaneously in cervical cancer. In the study by Kirchhoff et al. (1999) , none of the 29 invasive cancers analyzed showed an entire extra chromosome 3. A CEP3 probe was therefore used to evaluate the relationship between TERC amplification and polyploidy. A TERC: CEP3 ratio > 1 suggested isochromosome formation in a cell. Cells with a TERC: CEP3 ratio > 1 accounted for 61.0% of the cells in the normal/CIN1 group and 69.1% of the cells in the CIN2+ group. A TERC: CEP3 ratio of 1 was observed in 33.8% of the cells in the normal/CIN1 group and 26.9% of the cells in the CIN2+ group. The TERC: CEP3 ratio appears to be higher in the CIN2+ group than in the normal/CIN1 group, and the formation of an isochromosome provides a reasonable explanation for this observation.
In comparison with cytological analysis, the TERC test is suitable for CIN2+ diagnosis because of its high sensitivity (90.0%) and its optimal combination of sensitivity and specificity (Youden's index = 79.6). The C-MYC test is not suitable for cancer screening because its sensitivity and specificity are lower than those of the TERC test (80.0% vs. 90.0% and 77.7% vs. 89.6%, respectively). When we combined the C-MYC and TERC tests and considered one of the markers amplified to be positive, the sensitivity increased slightly from 90.0% to 92.0%, while the specificity decreased greatly from 89.6% to 72.0%. The C-MYC test showed marginally increased sensitivity for screening but reduced specificity.
The specimens used in the present study were primarily obtained from a population-based screening program in which residual cervical liquid-based specimens were used for the FISH test of TERC and C-MYC amplification and for the HC2 HPV DNA test. Patients with positive results (cytological analysis, TERC test, C-MYC test or HC2 HPV DNA test) were recommended for colposcopic examinations. Colposcopy-directed biopsy and histological evaluation were performed if indicated. The correlations among oncogene amplification, HPV infection and cytological-histological results were analyzed, and the design of an optimal strategy for cervical cancer screening is discussed in a separate article (Shaomin Chen, Yun Zhang, Yunbo Qiao, et al., manuscript in preparation). To analyze the correlation between oncogene amplification patterns and cytological/histological diagnosis in the present study, we chose all histologically confirmed cases, including 132 cases of NILM (including 84 HPV-positive cases and 48 HPV-negative cases; the 48 HPV-negative cases included 1 TERC+/C-MYC + case, 2 TERC+/C-MYC- cases, 10 TERC-/C-MYC + cases and 35 TERC-/C-MYC- cases; the 35 TERC-/C-MYC- cases were selected as a control group) and 55 cases of ASCUS+. This choice accounts for the higher number of histologically normal cases than abnormal cases. For the facilitation of the statistical analysis of amplification patterns, another 56 ASCUS + cases were recruited from February 2011 to October 2011. Although the number of abnormal cases is still rather small, the aberrant cells observed are sufficiently numerous for data analysis and it is possible to reach preliminary conclusions regarding the oncogene amplification patterns in CIN2+ cases. We collected detailed data (including contact information) from all of the patients who underwent screening for long-term follow-up. It is recommended that patients with negative screening results undergo routine screening and that patients with positive screening results undergo cytological analysis, FISH test, HPV test, colposcopy and histological evaluation at defined intervals.
The FISH test is suitable for clinical testing because of its several advantages. It can be performed using residual cytological specimens without additional sampling. The FISH test is a cell-based evaluation technique and is therefore more sensitive than other methods such as PCR and microarray-based analyses. The interpretation of fluorescent signals is objective and repeatable and does not rely heavily on highly trained personnel. We were interest in possible reasons for the discrepancy between FISH test results and cytological or histological results. We reviewed the slides of the 4 NILM/TERC-/CMYC- cases with underlying CIN2+ by screening the whole slide and found that the aberrant cell percentages were below the cut-off values and that the amplification patterns were simple (3-4 copies). We also reviewed the colposcopical and histological images from these cases and found that most of them were focal CIN2+ cases; therefore, sampling omissions could not be ruled out.
The preliminary results of this study indicate that TERC amplification is a clinically applicable genetic approach for cervical lesion diagnosis because of its high sensitivity and optimal combination of sensitivity and specificity. The C-MYC test cannot be used for screening because of its low sensitivity and because it does not result in increased specificity when used in combination with the TERC test. Compared to the C-MYC test, the TERC test shows more high-level amplification copies and more diverse amplification patterns in high-grade lesions. However, the sensitivity of the TERC test is lower when using a cut-off value for the GCN than for the cell percentage. Further investigation of the possible application of GCN for prognosis of cervical neoplasia is needed.