• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Gynecol Oncol. Author manuscript; available in PMC Jan 1, 2010.
Published in final edited form as:
PMCID: PMC2606919
NIHMSID: NIHMS75491

Early Invasive Cervical Cancer

MRI and CT Predictors of Lymphatic Metastases in the ACRIN 6651 / GOG 183 Intergroup Study
Donald G Mitchell, MD, FACR,a Bradley Snyder, MS,b Fergus Coakley, MD,c Caroline Reinhold, MD,d Gillian Thomas, MD, FRCPC,e Marco A. Amendola, MD, FACR,f Lawrence H Schwartz, MD,g Paula Woodward, MD,h Harpreet Pannu, MD,i Mostafa Atri, MD, FRCPC Dipl. Epid.,j and Hedvig Hricak, MD, PhDg

Abstract

Purpose

To compare MRI, CT, clinical exam and histopathological analysis for predicting lymph node involvement in women with cervical carcinoma, verified by lymphadenectomy.

Methods

A 25-center ACRIN/GOG study enrolled 208 patients with biopsy-proven invasive cervical cancer for MRI and CT prior to attempted curative radical hysterectomy. Each imaging study was interpreted prospectively by one onsite radiologist, and retrospectively by 4 independent offsite radiologists, all blinded to surgical, histopathological and other imaging findings. Likelihood of parametrial and uterine body involvement was rated on a 5-point scale. Tumor size measurements were attempted in 3 axes. Association with histologic lymph node involvement, scored as absent, pelvic only and common iliac or paraaortic, was evaluated using Cochran-Mantel Haenszel statistics, univariate and multivariate logistic regression, generalized estimating equations, accuracy statistics and ROC analysis.

Results

Lymphatic metastases were found in 34% of women; 13% had common iliac nodal metastases, and 9% had paraortic nodal metastases. Based on the retrospective multi-observer re-reads, average AUC for predicting histologic lymph node involvement between MRI and CT for tumor size were higher for MRI versus CT, although formal statistic comparisons could not be conducted. Multivariate analysis showed improved model fit incorporating predictors from MRI, but not CT, over and above the initial clinical and biopsy predictors, although the increase in discriminatory ability was not statistically significant.

Conclusion

MRI findings may help predict the presence of histologic lymph node involvement in women with early invasive cervical carcinoma, thus providing important prognostic information.

In women with cervical carcinoma that involves their lymph nodes, surgery alone is not sufficient treatment, and pelvic irradiation will not be curative if the tumor has metastasized to lymph nodes above the irradiated field. Unfortunately, even FDG PET/CT is not sensitive for detecting cervical carcinoma lymphatic metastases that have short axis diameter less than 5 mm.[1] Therefore, prognostic indicators are used to stratify patients based on their risk of having lymphatic metastases.[2-13]

Cross-sectional imaging tests such as CT and MRI are increasingly used to determine the extent of cervical carcinoma, often replacing components of traditional FIGO.[14-17] The recent American College of Radiology Imaging Network (ACRIN) / Gynecologic Oncology Group (GOG) multicenter clinical trial compared the performance of MRI, CT and FIGO clinical staging of invasive cervical cancer, verified by pathologic analysis of hysterectomy specimens.[18-20] Since analysis of hysterectomy specimens is not a perfect predictor of clinical outcome,[2, 4, 6] the principal aim of our current analysis is to evaluate MRI and CT, using the presence of lymph node metastases diagnosed at hysterectomy and lymphadenectomy (defined throughout this paper as histologic lymph node involvement) as a surrogate of poor clinical outcome among women referred for curative radical hysterectomy. Although final outcome is affected by postoperative adjuvant treatment, recurrence is more likely in women with lymphatic metastases.[4, 9, 21-32]

METHODS

Each imaging site was required to have a proven record of 20 surgical cases of cervical cancer per year, 1.5 T MRI and helical CT equipment, and an adequately qualified and committed radiologist, gynecologic oncologist, and pathologist. All institutions had study-specific institutional review board (IRB) approval. Between March 2000 and November 2002, 208 participants were accrued from 25 academic and community health centers. Methodology is described in further detail in earlier publications from this trial.[17, 18]

Participants

Consecutive participants with untreated biopsy-confirmed cervical cancer who were scheduled for curative hysterectomy based on pre-enrollment FIGO assessment were asked to participate. Imaging findings suspicious for metastatic involvement of lymph nodes (lymph node size greater than 1 cm in the short axis) were permitted to influence the decision to perform surgical biopsy or lymphadenectomy and potentially to cancel plans for radical hysterectomy. The interval between the first protocol imaging study and surgery could not exceed 6 weeks.

Data Acquisition and Analysis

All MRI and CT examinations met or exceeded standards agreed upon by the investigators. Technical parameters are described in greater detail in Hricak et al.[18]

No data were collected on women rejected for surgery on the basis of preoperative imaging findings, or on women who had retroperitoneal dissection only. All women had comprehensive pelvic lymph node dissection, but paraaortic dissection was performed at the discretion of the surgeon. Each surgeon completed a data form specifying the extent of disease found at surgery. Pathologists completed a similar form specifying presence or absence of malignancy in uterus (including lower uterine segment), parametrium, and lymph nodes in specific left and right anatomic regions, and measured the diameter of the primary tumor on the fixed tumor specimen.

Image Interpretation

One set of MRI and CT data forms were completed prospectively at each site by separate radiologist co-investigators, blinded to results of any other imaging test or clinical/pathology data. Images were then distributed digitally for retrospective multi-reader analysis by a group of eight experts in gynecologic oncology, four each for CT and MR imaging. Individual imaging findings relevant for staging were recorded using a 5-point scale (ranging from 1 = cancer definitely not present to 5 = cancer definitely present), including involvement of cervical stroma, vagina, uterus (lower uterine segment or higher), parametrium, and lymph nodes. If tumor margins could be delineated, tumor size was measured, in 3 axes if possible. The non-tumoral portion of cervix was not included in the diameter measurements. Mean tumor diameter was calculated for tumors when two or 3 diameter axes were measured. Tumors not seen were considered smaller than 1 cm.

Statistical Analysis

Three categories of histologic lymph node involvement were considered: negative (no involvement), low pelvic (not involving common iliac or above), and common iliac (with or without para-aortic). Additionally, binary responses were considered for common iliac histologic lymph node involvement (vs. low pelvic or no involvement), and any histologic lymph node involvement (vs. no involvement).

The association of histologic lymph node involvement and each clinical, pathological or imaging feature, was examined by means of Cochran-Mantel Haenszel (CMH) Statistics, employing the non-zero correlation statistic or the row mean score statistic, as appropriate, on the full ordinal scale. When possible exact p-values were computed and reported, otherwise the asymptotic chi-square approximation was used. These statistics are similar to the well-known chi-square test, but can have greater power to detect certain departures from the null hypothesis of row and column independence for ordinal variables. The Bonferroni procedure was used to account for the 36 multiple comparisons in the association analysis (12 prospective and 24 retrospective), with a P value of 0.0014 or less considered to indicate statistical significance. Additionally the strength and direction of the association was examined using univariate logistic regression models for the above described binary responses of general histologic lymph node involvement and histologic lymph node involvement specific to common iliac nodes, reporting odds ratios and associated 95% confidence intervals. Generalized estimating equations (GEE) were used for the multi-observer re-reads to account for the correlation induced by multiple readings of each case. For the multi-observer re-reads, agreement among readers for each feature was examined on the respective ordinal scale using the multi-rater kappa statistic [33]. Kappa (κ) values can be assessed as follows: 0.00 ≤κ< 0.40 indicates poor agreement; 0.40 ≤κ≤0.75, fair to good agreement; and κ> 0.75, excellent agreement [33].

A significant statistical association between a marker and the outcome of interest, as measured by the above statistics, does not always translate into high discrimination, or classification, accuracy [34]. Therefore, each feature’s ability to predict histologic lymph node involvement was also examined by means of the diagnostic probabilities of sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV), as well as by the area under the receiver operating characteristic (ROC) curve (AUC). In calculating ROC AUC, the full respective ordinal scale was employed for each predictor. For the retrospective reads, the average AUC for MRI and CT was compared using the method of DeLong et al. [35], and the Bonferroni procedure was used to account for the 4 multiple comparisons in this analysis, with a P value of 0.0125 or less considered to indicate statistical significance.

Lastly, a multivariate logistic regression model was employed based on the prospective readings to assess the additive predictive value of MRI and CT findings compared to the clinical assessment and biopsy alone, as follows: First, based on available clinical and biopsy predictors, selection procedures (forward, backward, and stepwise) were employed to help arrive at a parsimonious multivariate logistic regression model which adequately fit and explained the data. Goodness of fit was assessed via the Hosmer-Lemeshow test and adequacy of the reduced model was assessed via an appropriate likelihood ratio test. Once the baseline clinical model was determined, the MRI predictors of parametrial invasion, uterine involvement and average tumor size were added to the baseline clinical model and their joint significance tested via a likelihood ratio test. This step was then repeated separately for the CT predictors. For each modality, a formal comparison of the C-statistic (which is equivalent to the AUC derived using the predicted values from the model) between the baseline clinical model and the baseline clinical model plus imaging was also performed using the test statistic proposed by DeLong et al. [35]

RESULTS

A more detailed description of this patient cohort and multiobserver analysis is described in prior publications [18-20], and summarized in Appendix 1. Surgical and pathological tumor characteristics are summarized in Table 1. A representative case is illustrated in Figure 1.

Grade 3 poorly differentiated adenosquamous carcinoma with deep cervical invasion, left parametrial invasion, invasion of lower uterine segment, and positive paraaortic lymphadenopathy. Clinical assessment was FIGO stage IIA with estimated tumor diameter ...
Table 1
Summary of Histological and Surgical Findings

Surgico-Pathologic Lymph Node Data and Patient Outcome

Surgical and pathological data regarding histologic lymph node involvement were missing or incomplete in 11 of the 172 women, including the 3 who had aborted hysterectomy. Lymph node involvement was found pathologically in 55 of the remaining 161 women (34%). Of these 55 women with histologic lymph node involvement, 14 (25%) had common iliac and para-aortic involvement, 7 (13%) had common iliac but not paraortic involvement (high pelvis), and 34 (62%) had metastases restricted to the low pelvis.

N-staging by prospective imaging, scored as presence or absence of metastases at any site, showed 37% and 31% sensitivity, respectively, for both MRI and CT (not significantly different). Specificity was significantly higher for MRI (94%) than CT (86%), with a difference contrast of 0.09 (95% CI: 0.02, 0.15).

Tumor Size

Prospective readings

Average tumor diameter by pathology was less than 1 cm in 33 women (19%), 1-2 cm in 47 (27%), 2-3 cm in 32 (19%), 3-4 cm in 11 (6%), greater than 4 cm in 11 (6%), and unresolved or unavailable in 38 women (22%); these 38 women were excluded from analysis of tumor size by exam and imaging.

Tumor size could be measured by MRI in 153 (89%) women, compared with 126 (73%) by CT. Considering only cases where measurements were recorded, significant associations were detected between histologic lymph node involvement and average tumor size for MRI (p=0.0001) and clinical assessment (p=0.0002). After adjusting for multiple comparisons, no significant association was found for CT (p=0.007) or pathology (p=0.01). Estimated odds ratios, diagnostic probabilities and ROC AUCs are shown in Table 2.

Table 2
Odds ratios from univariate logistic regression models for average tumor size, using a threshold of >3 cm versus <=3 cm, by modality based on the prospective reads, along with the estimated diagnostic probabilities and ROC AUC for detection ...

Retrospective multi-observer readings

The multirater kappa statistic for CT was 0.35, and for MRI was 0.46. All data for the multi-observer readings are summarized in Table 3. There was no significant association between histologic lymph node involvement and average tumor size for any of the 4 CT readers (p=0.07, 0.50, 0.55, 0.73), and odds ratios were near unity for both general lymph node involvement and lymph node involvement specific to the common iliac nodes. After adjusting for multiple comparisons, one of the four MRI multi-observer re-readers demonstrated a significant association between histologic lymph node involvement and average tumor size (p=0.0005; 0.005, 0.01, 0.02). Odds ratios for average tumor size and corresponding 95% confidence intervals, along with average ROC AUC, for both MRI and CT are shown in table 3. Average AUC was higher for MRI than for CT (table 3), but could not be compared statistically because too few CT readers recorded measurements.[19]

Table 3
Odds ratios from univariate logistic regressions models employing generalized estimating equations for the features of average tumor size, parametrial invasion and uterine involvement by modality based on the retrospective multi-observer readings, along ...

Parametrial Extent

Prospective readings

By pathology, parametrial extent was bilateral in 6 (3%), unilateral in 9 (5%), absent in 123 (72%), and unknown or data unavailable by pathology in 34 (20%) women. A significant association was detected between histologic lymph node involvement and parametrial invasion, as detected by pathology (p=0.0011). After adjusting for multiple comparisons, the corresponding associations for MRI (p=0.02), CT (p=0.01) and clinical assessment (p=0.05) were not significant. Estimated odds ratios, accuracy statistics and ROC AUC were similar for each method for unilateral/bilateral invasion versus no invasion (Table 4), as well as for bilateral parametrial involvement versus unilateral or no parametrial involvement (results not shown).

Table 4
Odds ratios from univariate logistic regression models for parametrial invasion by modality based on the prospective reads, along with the estimated diagnostic probabilities and ROC AUC for detection of histologic lymph node involvement. Threshold was ...

Retrospective multi-observer readings

The multirater kappa statistic, averaged over left and right sides, was -0.04 for CT and 0.11 for MRI.[20] After adjusting for multiple comparisons, no significant associations between presence of histologic lymph node involvement and parametrial invasion (as determined by imaging) were detected for any of the 4 CT readers (p=0.006, 0.51, 0.71, 0.94). For MRI, a significant association was detected for 1 of the 4 readers (p=0.0002; 0.01, 0.05, 0.12). Odds ratios for parametrial invasion and corresponding 95% confidence intervals, along with average ROC AUC, for both MRI and CT are shown in Table 3. There was no significant difference in average AUC between MRI and CT for detection of histologic lymph node involvement specific to the common iliac nodes (p=0.10). Average AUC of parametrial involvement for detection of general histologic lymph node involvement was not significantly different for MRI versus CT, after adjusting for multiple comparisons (p=0.04).

Uterine Involvement

Prospective readings

Uterine involvement (lower uterine segment or higher) was present pathologically in 32 (19%), absent in 101 (59%), and unknown or data unavailable by pathology in 39 (23%) women. Histology was squamous in 19 of the 32 (59%) women with uterine involvement, compared with 73 of 101 (72%) without uterine involvement (p=0.17). Among tumors with available nonzero measurements, mean tumor size by pathology was 2.6 +/- 1.3 cm for the 28 tumors with versus 1.8 +/- 0.9 cm for the 82 tumors without uterine involvement (p=0.0046). The 32 women exhibiting uterine involvement included 7 of the 10 who had histologically involved common iliac nodes and available pathologic data (4/4 common iliac and paraaortic, 3/6 common iliac only). A significant association existed between histologic lymph node involvement and uterine involvement, as determined both by pathology (p=0.0001) and clinical assessment (p=0.0002).

Among women with available pathological material, prospective MRI readers considered the uterus definitely involved in 16, including 7 (50%) of the 14 women with histologically involved common iliac nodes, compared with 9 (6%) of the 147 who did not have involved common iliac nodes. CT prospective readers considered the uterus definitely involved in only 7 women overall, including 2 (14%) of 14 with involved common iliac nodes and 5 (3%) of 146 who did not have involved common iliac nodes. For both, MRI (p<0.0001) and CT (p=0.0004), a significant association was detected between histologic lymph node involvement and uterine involvement.

Estimated odds ratios, accuracy statistics and ROC AUC are shown in Table 5. The largest odds ratios for histologically involved common iliac nodes were observed for MRI and pathologic assessment, and the lowest for clinical assessment. The AUCs were similar for all modalities.

Table 5
Odds ratios from univariate logistic regression models for uterine involvement by modality based on the prospective reads, along with the estimated diagnostic probabilities and ROC AUC for detection of histologic lymph node involvement, using a threshold ...

Retrospective multi-observer re-reads

The multirater kappa statistic was -0.09 for CT and 0.10 for MRI. After adjusting for multiple comparisons, there was no significant association between histologic lymph node involvement and uterine involvement for any of the 4 CT readers (p=0.009, 0.01, 0.13, 0.90). A significant association was demonstrated for 1 of the 4 MRI readers (p=<0.0001, p=0.003, p=0.03, p=0.05). Odds ratios for uterine involvement and corresponding 95% confidence intervals, along with average ROC AUC, for both MRI and CT are shown in Table 3. None of the CT readers recorded a maximum grade of 5 (definitely present) for uterine involvement for any women with histologically involved common iliac nodes, and thus odds ratios could not be computed. The odds ratio for CT for general histologic lymph node involvement was close to unity. For MRI, women with a maximum grade of 5 for uterine involvement by MRI had an almost 6-fold increase in the odds of histologically involved common iliac nodes, and a more than 3-fold increase in the odds of histologic lymph node involvement in general (Table 3). However, there was no significant difference in average AUC between MRI and CT for detection of histologic lymph node involvement, either in general (p=0.63), or specific to the common iliac nodes (p=0.16).

Multi-variate analysis

The multivariate analysis of histologic lymph node involvement utilized the 161 cases with available lympadenopathy status. Histologically involved common iliac nodes could not be separately analyzed, as too few events resulted in model non-convergence.

The final baseline clinical model included the following covariates: tumor size by clinical exam, stromal invasion and cell type by histology, age, and race; as well as the following interactions: Age by race, age by clinical tumor size, and age by cell type. The likelihood ratio test compared with the full model was not significant at the 0.05 level (p=0.51), arguing for adequacy of the reduced model. Also, there was no demonstrable lack of fit at the 0.05 level (p=0.73). Thus, this was considered to be a parsimonious model based only on biopsy and clinical predictors which adequately fit and explained the data.

Adding the desired CT predictors of tumor size (>3 cm, <=3 cm), parametrial invasion (absent, unilateral/bilateral), and uterine involvement (5:Definitely present, <5) to the baseline clinical model demonstrated no improvement in model fit based on the likelihood ratio test (p=0.60), and no significant difference in the C-statistic (p=0.94).

Adding the same predictors as determined by MRI to the baseline clinical model did demonstrate a significant improvement in model fit based on the likelihood ratio test (p=0.01). Therefore, a multivariate logistic regression model consisting of biopsy and clinical plus MRI predictors was developed. The sole significant MRI predictor was tumor size, where women with average size >3 cm demonstrated greater odds of histologic lymph node involvement versus those with size <=3 cm, even after adjusting for biopsy and clinical predictors (OR 4.4; 95% CI: 1.8, 11.2). Additionally, the final model including clinical and MRI predictors showed good discriminatory ability (AUC=0.82; 95% CI: 0.75, 0.88). However, the increase in the C-statistic when MRI was added to the baseline clinical model was not significant (AUC increased from 0.79 to 0.82; p=0.21). These results show that the final model including MRI predictors leads to improved model fit, yet these MRI predictors were not shown to significantly increase the ability of the model to classify participants with or without the outcome compared to the clinical and biopsy predictors alone. However, given the improved model fit, of the two models in question, preference would be given to the final model including MRI predictors.

COMMENT

Most evaluations of cross-sectional imaging have assessed its accuracy compared with a pathologic reference standard. However, the radical surgical specimen is not available for preoperative clinical decision making, and the pathological findings are far from perfect in predicting clinical outcome. [2, 4, 6] It is therefore important to study diagnostic methods that can be utilized prior to and during clinical treatment, and to verify these methods based on clinical outcome whenever possible. For cervical cancer, predictors of lymphatic involvement are particularly important with regard to eventual outcome.

The ACRIN-6651/GOG-183 multicenter trial showed that MRI had higher agreement with pathology than CT or clinical examination for delineating tumor margins and measuring tumor size.[19] This trial also provides an opportunity to use histologic lymph node involvement as determined by radical hysterectomy and retroperitoneal lymphadenectomy tissue as an indicator of unexpected adverse clinical outcome. Of the 161 women with invasive cervical carcinoma treated for cure with radical hysterectomy and lymphadenectomy, unsuspected histologic lymph node involvement was found in 55 (34%); 21 (13%) had common iliac nodal metastases, and 14 (9%) had paraortic nodal metastases. Unfortunately, incomplete follow-up limited the value of our data regarding disease recurrence.

We found that pathology and MRI had similar ability to predict lymphatic metastases based on measurement of tumor size and uterine involvement. Our finding that less than half of tumors could be measured by CT retrospective readers due to poor delineation limits the potential utility of CT for measuring tumor size. The predictive ability of clinical exam for histologic lymph node involvement based on tumor size and parametrial involvement was similar to that of MRI, but MRI agrees more with pathologic measurement of tumor size, suggesting that MRI may have more potential value for treatment planning. [19] MRI is also more effective for determining uterine involvement [19]; evaluation of uterine extent may be a particular additive benefit of MRI. Analysis of these benefits of MRI versus its additive cost relative to clinical exam alone is beyond the scope of this investigation.

For MRI and CT, we were able to obtain multi-observer data in addition to the primary prospective readings, but could only evaluate the primary single-observer data for clinical and pathological assessment. Another important limitation was enrollment of women based on the expected availability of hysterectomy specimens for pathologic confirmation. This selection bias reduced the number of large or advanced tumors included, but our population of early cervical cancer is clinically important. Our ability to evaluate the association between parametrial involvement and histologic lymph node involvement was limited, since only 15 (9%) of women selected had parametrial involvement.

In conclusion, the ACRIN-6651/GOG-183 intergroup multicenter trial in women with early invasive cervical carcinoma showed unsuspected histologic lymph node involvement in 34% of radical hysterectomy specimens. Based on the retrospective multi-observer re-reads, after adjusting for multiple comparisons, there was no difference in average AUC for detection of histologic lymph node involvement between MRI and CT for parametrial invasion or uterine involvement. For tumor size, the estimates of average AUC were higher for MRI versus CT, although formal statistic comparisons could not be conducted. Lastly, we found improved model fit incorporating MRI predictors, but not CT, over and above the initial clinical and biopsy predictors. Our final multivariate model including MRI predictors demonstrated good discriminatory ability (AUC = 0.82), although the increase in discriminatory ability was not statistically significant (p=0.21). This may suggest that future studies to determine how imaging features could best be used to alter patient management based solely on FIGO criteria may be useful.

ACKNOWLEDGMENTS

This study was funded by NCI grant # U01 CA079778 and U01 CA080098 and was conducted jointly by the American College of Radiology Imaging Network (ACRIN) and the Gynecologic Oncology Group (GOG). Data were collected, managed and analyzed by the Biostatistics and Data Management Center of ACRIN. Both ACRIN and GOG participated in the study design, the data analysis, the writing of the report, and the decision to submit the paper for publication. The NCI did not participate in the study design, the collection, analysis, or interpretation of the data, the writing of the report, or the decision to submit it for publication.

Appendix 1

Patient Cohort and Data Related to the Primary Aim

Of the 208 participants enrolled, 9 (4%) were deemed ineligible because of enrollment disqualifications. Of the remaining 199 participants, 27 (14%) were excluded because of missing or incomplete data; thus 172 (83%) participants were included in the final analysis of prospective readings. Hysterectomy was abandoned in three of the 172 women, but detailed descriptions of surgical findings were provided. Lymphadenectomy was performed for 91% of women on the right, and for 92% on the left. There were no significant differences between the eligible and analysis sets of patients regarding demographics or clinical characteristics (data not shown).

Data Related to the Blinded Multi-observer Study

Separate analysis sets were defined for MRI and CT to maximize the sample size for each modality. For CT, 29 cases from the primary analysis set were excluded from the retrospective multi-observer study because of incomplete submission of digital image data. However, 3 cases excluded from the primary analysis set, due to lack of MRI imaging (n=2) or non-submission of local site CT interpretation data (n=1), had pathology reference standard information and could be included in the CT multi-observer analysis set, thus arriving at 73% (146/199) analyzable cases.

For MRI, 22 cases from the primary analysis set similarly were excluded in the multi-observer study, but 2 cases excluded in the primary analysis set could be included in the MRI multi-observer analysis set, thus arriving at 76% (152/199) analyzable cases.

REFERENCES

[1] Sironi S, Buda A, Picchio M, Perego P, Moreni R, Pellegrino A, et al. Lymph Node Metastasis in Patients with Clinical Early-Stage Cervical Cancer: Detection with Integrated FDG PET/CT. Radiology. 2006 Jan;238(1):272–9. [PubMed]
[2] Zreik TG, Chambers JT, Chambers SK. Parametrial involvement, regardless of nodal status: a poor prognostic factor for cervical cancer. Obstet Gynecol. 1996 May;87(5 Pt 1):741–6. [PubMed]
[3] Kosary CL. FIGO stage, histology, histologic grade, age and race as prognostic factors in determining survival for cancers of the female gynecological system: an analysis of 1973-87 SEER cases of cancers of the endometrium, cervix, ovary, vulva, and vagina. Semin Surg Oncol. 1994;10(1):31–46. [PubMed]
[4] Chung CK, Nahhas WA, Stryker JA, Curry SL, Abt AB, Mortel R. Analysis of factors contributing to treatment failures in stages IB and IIA carcinoma of the cervix. Am J Obstet Gynecol. 1980 Nov;138(5):550–6. [PubMed]
[5] Carl UM, Bahnsen J, Wiegel T. Radiation therapy of para-aortic lymph nodes in cancer of the uterine cervix. Acta Oncologica. 1993;32(1):63–7. [PubMed]
[6] Tsai CS, Lai CH, Wang CC, Chang JT, Chang TC, Tseng CJ, et al. The prognostic factors for patients with early cervical cancer treated by radical hysterectomy and postoperative radiotherapy. Gynecol Oncol. 1999 Dec;75(3):328–33. [PubMed]
[7] Huang HJ, Chang TC, Hong JH, Tseng CJ, Chou HH, Huang KG, et al. Prognostic value of age and histologic type in neoadjuvant chemotherapy plus radical surgery for bulky (>/=4 cm) stage IB and IIA cervical carcinoma. Int J Gynecol Cancer. 2003;13(2):204–11. [PubMed]
[8] Kodaira T, Fuwa N, Toita T, Nomoto Y, Kuzuya K, Tachibana H, et al. Comparison of prognostic value of MRI and FIGO stage among patients with cervical carcinoma treated with radiotherapy. Int J Radiat Oncol Biol Phys. 2003 Jul;56(3):769–77. [PubMed]
[9] Kim PY, Monk BJ, Chabra S, Burger RA, Vasilev SA, Manetta A, et al. Cervical cancer with paraaortic metastases: significance of residual paraaortic disease after surgical staging. Gynecol Oncol. 1998 Jun;69(3):243–7. [PubMed]
[10] Hong J, Lai C, Tsai C, Chang T, Wang C, Chou H. Recurrent squamous cell carcinoma of cervix after primary radiotherapy. Int J Radiat Oncol Biol Phys. 2003 Oct;57(2 Suppl):S344–5.
[11] Keys HM, Bundy BN, Stehman FB, Okagaki T, Gallup DG, Burnett AF, et al. Radiation therapy with and without extrafascial hysterectomy for bulky stage IB cervical carcinoma: a randomized trial of the Gynecologic Oncology Group. Gynecol Oncol. 2003 Jun;89(3):343–53. [PubMed]
[12] Mayr NA, Taoka T, Yuh WT, Denning LM, Zhen WK, Paulino AC, et al. Method and timing of tumor volume measurement for outcome prediction in cervical cancer using magnetic resonance imaging. Int J Radiat Oncol Biol Phys. 2002 Jan;52(1):14–22. [PubMed]
[13] Narayan K, McKenzie AF, Hicks RJ, Fisher R, Bernshaw D, Bau S. Relation between FIGO stage, primary tumor volume, and presence of lymph node metastases in cervical cancer patients referred for radiotherapy. 2003;13(5):657–63. [PubMed]
[14] Montana GS, Hanlon AL, Brickner TJ, Owen JB, Hanks GE, Ling CC, et al. Carcinoma of the cervix: patterns of care studies: review of 1978, 1983, and 1988-1989 surveys. Int J Radiat Oncol Biol Phys. 1995 Jul;32(5):1481–6. [PubMed]
[15] Chung H, Ahn HS, Kim YS, Lee EJ, Ryu HS, Chang KH, et al. The value of cystoscopy and intravenous urography after magnetic resonance imaging or computed tomography in the staging of cervical carcinoma. Yonsei Med J. 2001 Oct;42(5):527–31. [PubMed]
[16] Follen M, Levenback CF, Iyer RB, Grigsby PW, Boss EA, Delpassand ES, et al. Imaging in cervical cancer. Cancer. 2003 Nov;98(9 Suppl):2028–38. [PubMed]
[17] Amendola MA, Hricak H, Mitchell DG, Snyder B, Chi DS, Long HJ, 3rd, et al. Utilization of diagnostic studies in the pretreatment evaluation of invasive cervical cancer in the United States: results of intergroup protocol ACRIN 6651/GOG 183. J Clin Oncol. 2005 Oct 20;23(30):7454–9. [PubMed]
[18] Hricak H, Gatsonis C, Chi DS, Amendola MA, Brandt K, Schwartz LH, et al. Role of imaging in pretreatment evaluation of early invasive cervical cancer: results of the intergroup study ACRIN 6651-GOG 183. J Clin Oncol. 2005 Dec 20;23(36):9329–37. [PubMed]
[19] Mitchell DG, Snyder B, Coakley F, Reinhold C, Thomas G, Amendola M, et al. Early invasive cervical cancer: tumor delineation by magnetic resonance imaging, computed tomography, and clinical examination, verified by pathologic results, in the ACRIN 6651/GOG 183 Intergroup Study. J Clin Oncol. 2006 Dec 20;24(36):5687–94. [PubMed]
[20] Hricak H, Gatsonis C, Coakley FV, Snyder B, Reinhold C, Schwartz LH, et al. Early invasive cervical cancer: CT and MR imaging in preoperative evaluation - ACRIN/GOG comparative study of diagnostic performance and interobserver variability. Radiology. 2007 Nov;245(2):491–8. [PubMed]
[21] Ballon SC, Berman ML, Lagasse LD, Petrilli ES, Castaldo TW. Survival after extraperitoneal pelvic and paraaortic lymphadenectomy and radiation therapy in cervical carcinoma. Obstet Gynecol. 1981 Jan;57(1):90–5. [PubMed]
[22] Bleker OP, Ketting BW, van_Wayjen_Eecen B, Kloosterman GJ. The significance of microscopic involvement of the parametrium and/or pelvic lymph nodes in cervical cancer stages IB and IIA. Gynecol Oncol. 1983 Aug;16(1):56–62. [PubMed]
[23] LaPolla JP, Schlaerth JB, Gaddis O, Morrow CP. The influence of surgical staging on the evaluation and treatment of patients with cervical carcinoma. Gynecol Oncol. 1986 Jun;24(2):194–206. [PubMed]
[24] Delgado G, Bundy B, Zaino R, Sevin BU, Creasman WT, Major F. Prospective surgical-pathological study of disease-free interval in patients with stage IB squamous cell carcinoma of the cervix: a Gynecologic Oncology Group study. Gynecol Oncol. 1990 Sep;38(3):352–7. [PubMed]
[25] Alvarez RD, Potter ME, Soong SJ, Gay FL, Hatch KD, Partridge EE, et al. Rationale for using pathologic tumor dimensions and nodal status to subclassify surgically treated stage IB cervical cancer patients. Gynecol Oncol. 1991 Nov;43(2):108–12. [PubMed]
[26] Kataoka M, Kawamura M, Nishiyama Y, Hamamoto K, Hamada K, Iketani H, et al. Patterns of failure and survival in locally advanced carcinoma of the uterine cervix treated with high dose-rate intracavitary irradiation system. Int J Radiat Oncol Biol Phys. 1992;22(1):31–5. [PubMed]
[27] Landoni F, Maneo A, Colombo A, Placa F, Milani R, Perego P, et al. Randomised study of radical surgery versus radiotherapy for stage Ib-IIa cervical cancer. Lancet. 1997;350(9077):535–40. 1997/8/23. [PubMed]
[28] Hong JH, Tsai CS, Lai CH, Chang TC, Wang CC, Lee SP, et al. Postoperative low-pelvic irradiation for stage I-IIA cervical cancer patients with risk factors other than pelvic lymph node metastasis. Int J Radiat Oncol Biol Phys. 2002 Aug;53(5):1284–90. [PubMed]
[29] Trattner M, Graf AH, Lax S, Forstner R, Dandachi N, Haas J, et al. Prognostic factors in surgically treated stage ib-iib cervical carcinomas with special emphasis on the importance of tumor volume. Gynecol Oncol. 2001 Jul;82(1):11–6. [PubMed]
[30] Chou HH, Wang CC, Lai CH, Hong JH, Ng KK, Chang TC, et al. Isolated paraaortic lymph node recurrence after definitive irradiation for cervical carcinoma. Int J Radiat Oncol Biol Phys. 2001 Oct;51(2):442–8. [PubMed]
[31] Wang CJ, Lai CH, Huang HJ, Hong JH, Chou HH, Huang KG, et al. Recurrent cervical carcinoma after primary radical surgery. Am J Obstet Gynecol. 1999 Sep;181(3):518–24. [PubMed]
[32] Lai CH, Hong JH, Hsueh S, Ng KK, Chang TC, Tseng CJ, et al. Preoperative prognostic variables and the impact of postoperative adjuvant therapy on the outcomes of Stage IB or II cervical carcinoma patients with or without pelvic lymph node metastases: an analysis of 891 cases. Cancer. 1999 Apr;85(7):1537–46. [PubMed]
[33] Fleiss J, Levin B, Paik M. Statistical methods for rates and proportions. 3rd ed. Wiley-Interscience; Hoboken, NJ: 2003.
[34] Pepe MS, Janes H, Longton G, Leisenring W, Newcomb P. Limitations of the odds ratio in gauging the performance of a diagnostic, prognostic, or screening marker. Am J Epidemiol. 2004 May 1;159(9):882–90. [PubMed]
[35] DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988 Sep;44(3):837–45. [PubMed]
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

  • MedGen
    MedGen
    Related information in MedGen
  • PubMed
    PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...