Cervical cancer is a serious health problem accounting for 9 % of new cancer cases and 8 % of cancer deaths in women worldwide.In recent years, both overall survival and tumor-free survival rates have been significantly improved as a result of advances in the technologies of chemotherapy, surgery, and radiotherapy. Also, more attention has been directed to improving both the quality of life and psychosexual dysfunction resulting from these treatments.
Ovarian function preservation is critical for quality of life, especially for young patients (<45 years old). Preservation can be performed directly in situ or by ovarian transposition during surgery. Several risk factors for ovarian metastasis (OM) have been reported to facilitate the decision to preserve the ovaries during surgery. These include large tumor size, lymphovascular space involvement (LVSI), lymph node metastasis (LNM), deep stromal invasion (DSI), parametrial invasion (PMI), corpus uteri invasion (CUI), and histology.However, there is still no consensus on the selection criteria for ovarian preservation because of the small number of studies and sample size restrictions.The study performed by Shimada et al. reported as many as 52 patients with OM out of 3,471 patients with International Federation of Gynecology and Obstetrics (FIGO) stages IB to IIB cervical cancer; however, there were no detailed statistical analyses of risk factors for OM.
Neoadjuvant chemotherapy (NACT) before surgery is widely used in many countries, such as Asia, Italy and South America, because of its potential for increasing resectability and reducing micrometastases.
In our country, China, NACT plus radical surgery has been used for patients with FIGO stage IB1–IIB cervical cancer for several years. Recently, NACT has been extended to the application of fertility-sparing surgery, combined with the advantage of reducing tumor volume. It was reported that NACT may reduce prognostic risk factors (LNM, PMI, LVSI, DSI, and large tumor size), which were also the potential risk factors for OM.However, until now, no research has used a large enough patient population to determine the influence of NACT on OM. Therefore, we designed this multicenter retrospective study to calculate the effects of NACT and its combined effects with the potential risk factors for OM. We also combined all the relevant published data to perform a meta-analysis on the pooled estimated effects of the reported risk factors for OM.
In this retrospective case-control study, clinical data from 10,897 patients were collected from the cervical cancer database, v1.10 (http://clinicaltrials.gov; NCT01267851). Cervical cancer patients with FIGO stage IB1–IIB disease were enrolled onto this study from January 2002 to December 2008, including squamous cell carcinoma (SCC), adenocarcinoma, adenosquamous carcinoma, and small cell carcinoma of the cervix. This study was reviewed by the ethics committee of Huazhong University of Science and Technology, and informed consent was obtained from all patients.
At our institutions, young patients (<45 years old) and patients older than 45 with regular menstrual cycles were generally offered ovarian preservation if the following were not present: stage IIB or later disease; adenocarcinoma, adenosquamous carcinoma, or small cell carcinoma; suspicious LNM; suspicious CUI; or abnormal-appearing ovaries. If they had any one of these clinicopathologic factors, we suggested that they undergo radical hysterectomy, pelvic lymphadenectomy, and bilateral salpingo-oophorectomy. The decision to preserve one or both ovaries also depended on patient preference. The characteristics of patients with or without ovarian preservation are provided in Supplemental Table S1.
To exactly evaluate the associations between clinicopathologic factors and OM by pathological examination of the operation specimen, all enrolled patients underwent a radical hysterectomy, pelvic lymphadenectomy, and bilateral salpingo-oophorectomy with or without NACT. NACT was generally administered in one or two courses depending on tolerance and response, and only patients with good response received an additional one or two cycles. The NACT regimens used in our study are provided in Supplemental Table S2.
OM was defined as the occurrence of viable tumor cells in the ovarian tissues or vessels. The following patients were excluded from analysis: patients who had other primary malignant tumors simultaneously (one patient had an endometrial carcinoma, and another had an ovarian carcinoma), patients whose lesion’s primary site could not be distinguished from that of the corpus uteri, and patients whose ovarian tumor’s histopathologic type did not accord with that of the cervical lesion.
The clinical and pathological data, including age at diagnosis, FIGO stage, treatment modality, follow-up, histological type, tumor size (≤4 cm or>4 cm), DSI (≤2/3 or>2/3 depth), LVSI, LNM, PMI, and CUI, were collected. A follow-up examination for patients with OM was recommended every 3 months for the first year, then every 6 months for the next 4 years. The median follow-up period was 34 months, with a range of 3–68 months.
The Pearson χ2 test or the Fisher exact test was used to assess the association between the clinicopathologic features of cervical cancer and the presence of OM. In the multivariate analysis, a binary logistic regression model was used to calculate the associations of independent risk factors. In order to provide more power for the selection of variables with relatively weak effects, the variables with P<0.2 in univariate analysis and the variables that were reported to be associated with OM in the literature were selected into the logistic regression analysis.13 NACT may have a protective effect on OM and thus was also selected in the multivariate analysis. Stepwise backward selection procedures (P<0.10 to enter the study and P<0.05 to remain in the study) were used in logistic regression analysis. All statistical tests were two-sided, and P values of less than 0.05 were considered significant. Analyses of case-control data were performed by SPSS software 13.0 (SPSS, Chicago, IL).
Medline was searched (last search performed on August 1, 2011) with the following keywords: cervical cancer, ovarian metastasis, risk factors. Bibliographies of the retrieved articles were examined for any further relevant publications. Only studies published in English and with a case-control design that contained data sufficient to calculate an odds ratio (OR) and the 95 % confidence interval (CI) were included in the meta-analysis. If a case-control design was not used, if there were no extractable data, or if risk factors were reported without being calculated, the articles were excluded from the meta-analysis.
Heterogeneity was evaluated using the Q statistical test (significant for P<0.1). A continuity correction of 0.5 was added to the cell frequency in studies with a zero-count outcome. Pooled ORs for pathological risk factors thought to be closely related to OM, such as LNM, CUI, DSI, histological type, PMI, LVSI, and large tumor size, were calculated using a fixed-effect or random-effect model according to the heterogeneity between the studies. Metaanalyses were performed using the metabin option from the Meta package of the R statistical software (http://www. R-project.org).
A total of 1,889 patients from the cervical cancer database v1.10 were enrolled onto the present study (Fig. S1). The distributions of histological types were as follows: 1,621 patients (85.8 %) with SCCs, 183 patients (9.7 %) with adenocarcinomas, 72 patients (3.8 %) with adenosquamous carcinomas, and 13 patients (0.7 %) with small cell carcinomas. A total of 1,030 patients (54.5 %) were diagnosed with stage IB, 561 patients (29.7 %) with stage IIA, and 298 patients (15.8 %) with stage IIB disease. The mean age of all patients was 48.7 (range 22–81) years. The characteristics of patients treated with primary surgery or NACT plus radical surgery are shown in Supplemental Table S3.
Twenty-two patients (1.2 %) were diagnosed as OM (Table 1). The mean age of these 22 patients was 46.3 (range 33–62) years. The ovarian metastatic rates of the four histological types were 0.7 % (12 of 1621), 2.7 % (5 of 183), 5.6 % (4 of 72), and 7.7 % (1 of 13) for SCCs, adenocarcinomas, adenosquamous carcinomas, and small cell carcinomas, respectively, and there was a significant ascending trend (linear by linear association test, P<0.001) for ovarian metastatic rates with histology (SCCs, adenocarcinomas, adenosquamous carcinomas, and small cell carcinomas). More detailed information about ovarian metastatic rates with respect to the histological types and FIGO stages is shown in Table 2.
Potential risk factors for OM are listed in Table 3. Patients with OM were frequently observed with LNM (P<0.001), CUI (P<0.001), PMI (P<0.001), non-SCC (including adenocarcinoma, adenosquamous carcinoma, and small cell carcinoma; P<0.001), DSI (P = 0.002), or higher FIGO stage (P = 0.03) in univariate analysis. The ovarian metastatic rate was slightly lower in patients who received NACT plus radical surgery, and there was no statistically significant difference (0.8 vs. 1.3 %, P = 0.36). The following variables were selected together with NACT and variables with P<0.2 (in univariate analysis) into the logistic regression analysis: LVSI (P = 0.52) and tumor size (P = 0.30). Stepwise regression analysis identified LNM (OR 5.75, 95 % CI 2.16–15.28), CUI (OR 5.53, 95 % CI 2.11–14.53), PMI (OR 8.24, 95 % CI 3.01–22.56), and histology and NACT (OR 0.40, 95 % CI 0.13–1.22) to be independently associated with OM.
Furthermore, we also constructed two logistic regression models using the data of patients with SCC and adenocarcinoma (Table 4), respectively. Multivariate analysis revealed that OM was associated with PMI (OR 5.67, 95 % CI 1.63–19.72), CUI (OR 3.25, 95 % CI 0.88–12.01), and LNM (OR 9.44, 95 % CI 2.43–36.65) in patients with SCC. NACT had no significant protective effects (OR 0.64, 95 % CI 0.15–2.69, P = 0.55) on OM in patients with SCC after adjustment of other clinicopathologic variables (including age, tumor size, LNM, CUI, PMI, LVSI, DSI, and FIGO stage). However, FIGO stage (IIB vs. IB, OR 31.78, 95 % CI 1.41–716.33), bulky tumor size (OR 12.71, 95 % CI 1.31–123.68), PMI (OR 51.21, 95 % CI 4.10–639.19), NACT (OR 0.003, 95 % CI 0.00–0.27), and CUI (OR 44.49, 95 % CI 2.77–714.70) were independent risk or protective factors for OM in patients with adenocarcinoma.
On the basis of our criteria, a total of 11 published articles from the literature were identified. We excluded irrelevant articles, case reports, or articles not published in English. Each publication was evaluated by two independent researchers. Out of the 11 papers, only nine met our criteria, which provided sufficient data to calculate the pooled OR.The DSI category in the study by Kim et al.had slight differences compared to the present study. However, to maximize the statistical power, it was not excluded from our meta-analysis. In Fig. 1a, b, the following clinicopathologic factors were closely related to OM: LNM (OR 7.89, 95 % CI 5.07–12.26), CUI (OR 6.59, 95 % CI 3.10–14.04), LVSI (OR 3.35, 95 % CI 1.71–6.60), PMI (OR 5.68, 95 % CI 2.61–12.33), adenocarcinoma vs. SCC (OR 5.49, 95 % CI 3.23–9.33). In addition, DSI was a risk factor with marginally significant overall effect on OM (OR 3.07, 95 % CI 0.99–9.56).
In our database, 6,400 (58.7 %) of 10,897 of patients were diagnosed as having cervical cancer before the age of 45. Ovarian preservation surgery may be desired by these women of reproductive age; however, some cases with OM have been reported, and the safety of this kind of treatment remains debatable. Until now, no criteria have been defined for ovarian preservation in cervical cancer.
Histological type was an important variable affecting OM in cervical cancer from the results of our case-control study and meta-analyses. In the present study, the incidence of OM from adenocarcinoma was significantly higher than from SCC. Nakanishi et al. reported that the differences in OM between SCC and adenocarcinoma may be due to the different characteristics of these two types of carcinomas.Our multivariate analysis revealed that LNM was an independent risk factor for OM in SCC not adenocarcinoma, and large tumor size, FIGO stage, and NACT were significantly associated with OM in adenocarcinoma not SCC. It was revealed that there may be some differences in metastatic routes of OM between these two types of carcinomas.
In the literature, lymphatic metastasis is considered to be the most important route for OM.In this study, we identified LNM as an independent risk factor for OM in SCC. In the further analysis, LNM was also identified to be independent risk factor with the largest effect for OM in SCC not in adenocarcinoma. In only one case of nine were the involved ovary and the involved pelvic lymph node in SCC patients with OM on the same side. Therefore, it seems that lymphatic metastasis is the most important metastatic route for OM in SCC.
Hematogenous spread was reported to be the potential route for OM.Tabata et al. reported that adenocarcinoma tended to metastasize to the ovary even without LNM, and the incidence of OM in this histological type was not related to the number of metastasized lymph nodes.In our study, the critical role of hematogenous spread rather than lymphatic metastasis in adenocarcinoma can be supported by the protective effect of NACT against OM, which eliminated potential micrometastasis in the bloodstream, and the insignificant effect of LNM on OM in adenocarcinomas after adjustment. Therefore, from the view of different metastatic routes between SCC and adenocarcinoma, the criteria for ovarian preservation of these two histological types should be established separately.
As early as in 1987, Tabata et al.pointed out that OM easily occurred when it invades the uterine corpus. In the past two decades, lymph node metastases, DSI, and LVSI have also been reported to be associated with OM in cervical cancer.To identify the most important risk factors for OM, many researchers have tried to perform multivariate analysis of large data sets. Sakuragi et al. found that blood vessel invasion and PMI were significant independent risk factors for OM. However, Yamamoto et al. reported that OM was significantly associated with histologic type and blood vessel invasion in multivariate analysis. In addition, a retrospective study by Landoni et al.found only age, FIGO stage, histology, and unaffected peripheral stromal thickness to be independent risk factors for ovarian metastases. Perhaps because of the relatively small sample size and the heterogeneity of enrolled patients, these findings were so different from each other that there was still no consensus on the criteria for ovarian preservation in the surgical treatment of cervical cancer.
In our study, we pooled all the data from previously published case-control studies by meta-analyses and found that LNM, CUI, and PMI had the greatest effects on OM in cervical cancer. Furthermore, these three pathological variables were also the independent risk factors for OM in SCC patients from our case-control study, and two of them (CUI and PMI) were independently associated with OM in adenocarcinomas. In our multivariate analysis, OM was associated with PMI, CUI, and LNM in patients with SCC, whereas FIGO stage (IIB vs. IB), bulky tumor size, PMI, NACT, and CUI were independent risk or protective factors for OM in patients with adenocarcinoma. Therefore, the present study suggested that ovarian preservation should not be performed in SCC patients with suspicious LNM, CUI, or PMI, or in patients with adenocarcinomas with FIGO stage IIB, bulky tumor size, suspicious CUI, or PMI. DSI and LVSI were identified to be risk factors for OM in our univariate analysis from case-control study or meta-analyses with mild to moderate effects, but not to be independent risk factors for OM in our case–control study. We thus suggest that, under regular follow-up examinations for ovarian disease recurrence, ovarian preservation also could be performed in patients with DSI or LVSI.
NACT was initially used in locally advanced cervical cancer patients, such as FIGO stage IB2 and IIB disease, and tumor-size reduction was an important rational for the use of NACT.In recent years, NACT was also applied to earlier cervical cancers (stage IB–IIA<4 cm) in our country as well as others.Because there is no consensus on the survival benefit of NACT followed by radical surgery, it still has not been recognized as a standard treatment for cervical cancer patients in many countries. However, as reported in our previous study, some patients could obtain benefit from NACT, such as patients whose disease responds clinically, and it is important to establish selective criteria for patients who would benefit most from NACT.
Until now, the influence of NACT on OM has not been reported. Our multivariate analysis demonstrated that NACT was an independent protective factor against OM in adenocarcinoma. Given that the incidence of OM in adenocarcinoma was much higher than in SCC, many researchers recommended that ovarian preservation should not be performed in any patients with cervical adenocarcinoma.However, in the present study, the incidence of OM was only 1.6 % (1 of 64) in adenocarcinomas treated with NACT plus radical surgery. It seems that ovary preservation may be safe in adenocarcinomas if the patient underwent preoperative chemotherapy. Therefore, this protective effect against OM is an attractive advantage for NACT plus radical surgery for patients with cervical adenocarcinoma compared to primary radical surgery, and this may become a new rational for the use of NACT. In addition, NACT provides opportunities for fertility-sparing treatments for reproductive-aged women with a tumor size of>2 cm in FIGO stage IB–IIA cervical cancer. Thus, the protective effects of NACT against OM also provide beneficial evidence for this treatment modality.
Small cell carcinoma of the cervix is a rare and aggressive tumor, the incidence of which is secondary to both SCC and adenocarcinomas.In 1993, Young et al. reported two cases of small cell carcinoma with ovarian involvement.However, the incidence of OM from this type of cervical cancer had not been recognized. In the present study, the incidence of OM from small cell carcinomas was 7.7 %, which was significantly higher than that observed in both SCC and adenocarcinomas. Therefore, we do not suggest preserving ovaries in this rare type of cervical carcinomas.
There are some limitations to our study. First, this was a multicenter retrospective study, and selection bias inherent to the kind of design may be introduced in our study. The best way to avoid or minimize this bias is to conduct a prospective randomized study. However, as Landoni et al. mentioned, a prospective randomized study for this purpose is not feasible for ethical reasons and because of the high number of cervical cancer patients who should be enrolled to show a statistically significant difference.Retrospective study with large data sets is still the best way to investigate these issues. Second, OM is a rare disease, and thus the sample size of patients was small in our casecontrol study, especially the patients treated with NACT. Despite these limitations, to our knowledge, this is the first study to investigate the potential role of NACT in ovarian preservation. Moreover, using meta-analysis, we pooled all related studies to evaluate potential risk factors for OM with a large enough overall sample size.
In conclusion, on the basis of the present data and metaanalyses, ovarian preservation surgery may be safe in SCC patients without suspicious LNM, PMI, and CUI and in adenocarcinomas receiving NACT without FIGO stage IIB, large tumor size, suspicious PMI, and CUI. In addition, ovarian preservation should not be performed in patients with small cell carcinoma.