Page 191 - Strategies for non-invasive managementof high-grade cervical intraepithelial neoplasia - prognostic biomarkers and immunotherapy Margot Maria Koeneman

As such, clinical application of 3q26 gain, as a single marker, is limited by suboptimal test characteristics.
The results of our biomarker studies emphasize another important challenge in prognostic biomarker research in high-grade CIN: the achievement of adequate prognostic test properties. Desirable performance levels depend on the clinical context of the biomarker.[24] In case of high- grade CIN prognosis, it is vital to identify those lesions that will not regress spontaneously, as these women require treatment to avoid malignant evolution. Consequently, a high specificity of a test in prediction of disease regression is warranted. At the same time, a high sensitivity limits overtreatment, by correct identification of those lesions that will regress spontaneously. In our review, we concluded that no individual biomarker showed sufficient prognostic value to be implemented as a single prognostic test. This may be explained by the fact that individual biomarkers reflect contributing, but not single critical steps, in cervical oncogenesis.
The development of CIN and cervical cancer is characterized by a complex interaction between virus and host, in which viral oncogenic properties and the human immune system influence the cellular processes that lead to cervical oncogenesis.[25, 26] In this process, several important molecular events have been identified, among which are for example inactivation of pRB and p53, viral DNA integration and upregulation of telomerase.[25] However, no individual event has been identified as ‘point of no return’ in disease progression. Indeed, viral DNA integration and upregulation of telomerase is not found in all high-grade CIN lesions or cervical carcinomas and lesions with low pRB/p53 can still regress.[27-29] As such, it is unlikely that the prediction of the natural prognosis of CIN lesions will be based on a marker reflecting one molecular event, but rather on a combination of viral, host and genetic parameters. Therefore, in order to achieve better test performances, we suggest the development of biomarker profiles or prediction models, instead of aiming for individual prognostic markers. Previous research has shown that biomarker combinations improve test performance, when an individual marker does not have sufficient classification accuracy on its own.[30] One such biomarker profile in high-grade CIN has previously been described.[11] This profile was developed in 162 women with high-grade CIN and consists of four markers: the percentage of pRb in the lower epithelium, lesion size, the amount of CD4+ cells in the stroma and consistent condom use. All four markers are independently associated with disease regression. The model categorizes women into a group with a low regression probability of 8% and a high regression probability group of 56%. The model was characterized by a sensitivity of 71% and a specificity of 85%: 24 out of 34 regressive lesions and 109/128 non-regressive lesions were correctly predicted. This implies that overtreatment would occur in 30% (10/34) of women and undertreatment in 15% of women (19/128). As such, the model seems insufficient for implementation in clinical practice, but serves as an example for future research. Biomarker profiles composed of other markers may provide better test characteristics.
Prognostic markers for spontaneous regression of CIN2, as a subset of high-grade CIN
An important consideration regarding observational studies in high-grade CIN is the distinction between CIN2 and CIN3 lesions, because spontaneous regression of CIN2 seems much more common than CIN3. Natural history studies, in which CIN2 and CIN3 lesions were distinguished
10
Discussion
189