Tumor Marker Score for Prognostication of Early-stage Squamous Cell Cervical Cancer

Discussion

A score using tumor marker expression for prognostication in cervical cancer has not been previously assessed. The score presented here shows a strong correlation to 10-year mortality in early-stage cervical cancer. The score is, however, not solely intended to direct treatment, but to serve as an adjunct to clinical and histopathological findings. Such a score will aid in identifying women with early-stage cancer who would benefit from aggressive treatment, i.e. combined treatments, and those who would probably not. A score for late-stage cancer might be of less interest, as these patients have a very poor prognosis and will receive aggressive therapy.

Sensitivity, specificity, PPV and NPV give information on whether the score is clinically-useful. Thus, the high specificity and NPV when scores 0-2 and 3-4 were compared is satisfactory as women with good prognosis will not have unnecessary aggressive treatment. The sensitivity and PPV were lower and suggest that a low score could also be associated with mortality and that in fact a minority of these women would benefit from aggressive treatment. We believe, however, that sensitivity and PPV are acceptable, as the score should be used as an adjunct to clinical prognostic factors, such as specified growth, age, positive lymph nodes and general condition of the patient. Taken together, these factors may define the large majority of women with poor prognosis despite low scores.

There is a basis for excluding women with score 2, as their mortality is close to the overall mortality. Hence, when score 0-1 and 3-4 were compared, estimations gave a very high sensitivity that would leave few women with a poor prognosis not treated aggressively. The NPV was similarly high, i.e. a low score will also reflect a favorable prognosis. Specificity was, however, slightly lower, compared to comparisons of score 0-2 vs. 3-4. Exclusion of the intermediate score seems to be the most attractive model, but should be evaluated in further studies.

The selection of tumor markers in this study serves as a model. Different local traditions, study populations, laboratories, clones and antibodies will influence tumor marker expression. An optimal prognostication score might include our investigated, and other, tumor markers. The tumor markers included here are all biologically-plausible for prognostication and have previously been investigated as single prognostic markers. In the present study, we were able to use our experience from three previous studies to construct a score from five tumor markers that turned out to be clinically useful.

We previously presented the crude correlations with prognosis for 10 out of the 12 tumor markers included in the present study (13). p53, c-MYC, COX2 and CD4 expression appeared promising, and have been widely studied for prognosis in cervical cancer (14-18). In our previous studies, all cancer stages were included, although adjustment for stage was made in multivariate analysis. Models with combinations of tumor marker expression improved prognostication, but were in general clinically unsatisfactory. Subsequently, we were able to evaluate the novel protein LRIG1, and finally LRIG2. In early-stage, but not late-stage cancer LRIG1 expression appeared to correlate to favorable (13), and LRIG2 expression to poor, prognosis (19).

c-MYC is one of the ‘classic’ oncogenes, and its translocation in Burkitt's lymphoma was first shown in 1982 (20). The functions of c-MYC products are still not completely understood as they bind to hundreds of potential target genes. It is however evident that c-MYC expression contributes to increased proliferation and loss of differentiation in cancer (21). Previous studies have concluded that c-MYC expression is increased and is involved in cervical cancer (17, 18). In our model, c-MYC expression did not significantly correlate to a poor prognosis, but the p-value was borderline, and its inclusion was useful.

p53 is a major tumor suppressor and its expression has been studied in a majority of solid cancer types. p53 induces cell-cycle arrest at the G₁ and G₂ checkpoints prior to DNA replication, allowing repair of damaged DNA, but also induction of apoptosis, thereby hampering development of cancer cells (22). In cervical cancer, the human papillomavirus oncogene E6 is able to promote p53 degradation (23). In contrast, in most cancer types, inactive mutant p53 is found. Results of evaluating mutant p53 in cervical cancer have, therefore, been confusing. In the present study, expression of active wild-type p53 was evaluated (16).

COX2 was included in the study because of its correlation to inflammatory response. COX2 is involved in a number of negative events in cancer, such as angiogenesis and reduced apoptosis (9). COX2 expression has been correlated to poor prognosis in many previous studies on cervical cancer (24).

The LRIG1 gene is located on chromosome 3p14.3. This region is frequently deleted in cancer. In addition to our study, a correlation with a good prognosis in skin cancer has been reported. LRIG1 has also been shown to be an important determinant of cancer growth and good prognosis in prostate cancer (10), brain tumors (25) and in mediating receptor degradation in breast cancer (26). LRIG1, thus, has a role as a tumor suppressor in a variety of solid, epithelial cancer types (11, 13).

Expression of another member of the LRIG family, LRIG2, in contrast, was recently reported to be strongly associated with glioma risk (27) and poor prognosis in oligodendroglioma (25, 28), as well as in our study with early-stage cervical cancer (19).

All previous scoring models have, with one exception (8), included lymph node metastasis. This necessitates surgery, either explorative, or as treatment. Radiotherapy and radical hysterectomy are both major treatment options for early-stage cancer. When radiotherapy is the treatment of choice, and surgical exploration is not routinely performed, these histopathological scores are thus not useful (4, 5, 7, 9, 10, 29-31). In addition to lymph node status, factors such as tumor size, histopathological type, depth of invasion, lymph vascular invasion, parametrial extension and radical surgical margins have been evaluated. All studies reported that their score correlated to survival. In no study, however, were the sensitivity, specificity, PPV and NPV estimated. Statistical analyses were restricted to p-values, risk ratios and Kaplan–Meier curves, which is why it is difficult to evaluate their clinical usefulness.

The present study has presented a novel prognostication score, aimed as an adjunct to clinical and histopathological parameters in choosing treatment of early-stage squamous carcinoma of the uterine cervix. This score is assessable on limited biopsy material before any treatment, whether radiotherapy, surgery or adjunct chemotherapy, and with adequate sensitivity, specificity, PPV and NPV may valuably support decision-making to avoid unnecessary aggressive treatment in women with a favorable prognosis, and to identify women with a poor prognosis requiring for more aggressive therapy.