| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
ORIGINAL RESEARCH |
From the Departments of 1Gynecology and Obstetrics and 3Oncology, Centre Hospitalier Intercommunal de Créteil, University Paris 12, Créteil, France; and 2Department of Gynecology and Obstetrics, University of Torino, Torino, Italy.
 |
ABSTRACT |
|---|
TOP
 ABSTRACT MATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
METHODS: Data from 244 patients treated for vulvar cancer at a single institution (Creteil, France) were used as a training set to develop and calibrate a nomogram for predicting relapse-free survival and local relapse-free survival. We used bootstrap resampling for the internal validation and we tested the nomogram on an independent validation set of patients (Torino, Italy) for the external validation.
RESULTS: The nomograms were based on a Cox proportional hazards regression model. Covariates for the relapse-free survival model included age, T stage, number of metastatic nodes, bilateral lymph node involvement, omission of the lymphadenectomy, margin status, lymphovascular space invasion, and depth of invasion. The concordance indices were 0.85 and 0.83 in the training set before and after bootstrapping, respectively, and 0.83 in the validation set. The predictions of our nomogram discriminated better than did the International Federation of Gynecology and Obstetrics stage (0.83 compared with 0.78, P = .01). The calibration of our nomogram was good. In the validation set, 2-year and 5-year relapse-free survival were well predicted with less than 5% difference between the predicted and observed survivals for each quartile. A nomogram for predicting local relapse was also developed.
CONCLUSION: We have developed nomograms for predicting distant and local relapse of vulvar cancer at 2 and 5 years and validated them both internally and externally. These nomograms will be freely available on the International Society for the Study of Vulvovaginal Disease Web site.
LEVEL OF EVIDENCE: III
|
MATERIALS AND METHODS |
|---|
|
TOP
ABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
Relapse-free survival and local relapse-free survival were estimated using the Kaplan-Meier method. We defined relapse as inguinofemoral, pelvic, or distant recurrences of the disease, whereas local relapse was considered separately to develop a specific nomogram. We used the Cox proportional hazards regression model for the multivariate analyses and for the construction of the nomograms. Backward variable selection was performed using a method based on Lawless and Singhal7 to determine independent covariates. Continuous variables were fit using restricted cubic splines to relax the linearity assumptions if necessary.8 The model performance was quantified with respect to discrimination and calibration. Discrimination, ie, whether the relative ranking of individual predictions is in the correct order, was quantified with the concordance index, which is similar to the area under the receiver operating characteristic curve but appropriate for censored data. The concordance index is the probability that given 2 randomly selected patients, the patient with the worse outcome will in fact have a worse outcome prediction. The concordance index ranges from 0 to 1, with 1 indicating perfect concordance, 0.5 indicating no association (no better than flipping a coin), and 0 indicating perfect discordance. We used the bootstrapping technique to obtain relatively unbiased estimates (200 repetitions). Calibration, ie, agreement between observed outcome frequencies and predicted probabilities, was studied with graphic representations of the relationship between the observed outcome frequencies and the predicted probabilities (calibration curves) for groups of patients defined by quartiles (each quartile contained at least 20 cases). The nomogram was also validated externally using the Torino series by comparing nomogram predictions with the observed rates. This was done by grouping patients with respect to their nomogram-predicted probabilities and then plotting the mean of the group with the observed Kaplan-Meier estimate of disease-specific survival. All analyses were performed using the R package with the Design, Hmisc, and Lexis libraries (http://lib.stat.cmu.edu/R/CRAN/).
|
RESULTS |
|---|
|
TOP
ABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
|
|
In the training set, the local relapse rates were 87% and 78% at 2 and 5 years, respectively. Margin status (P = .0002), depth of invasion (P = .001) and lymphovascular involvement (P = .01) were independently associated with local relapse-free survival in the multivariate analysis. A nomogram on the basis of this Cox model appears in Figure 1B. The predictions against the actual outcome for the external validation set are reported in Figure 3: 2-year and 5-year local relapse-free survival were both well predicted. However, the 5-year local relapse rate was slightly overestimated for patients at high risk of local relapse. In this independent set, the concordance index of the nomogram was 0.76.
|
We developed a computer program to provide a more friendly and useful version of our nomogram. This program may help patients and physicians with the difficult decision-making about the need for integrated therapies. The program, called Vulvarcancer! (version 1), is programmed in Java text. An Internet browser with Java enabling is required to run the applets. The applets are available online on the International Society for the Study of Vulvovaginal Disease Web site.
|
DISCUSSION |
|---|
|
TOP
ABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Clinical consensus has long held that the absolute number of positive inguinofemoral lymph nodes is the most important prognostic factor in vulvar cancer. Somewhat surprisingly, this consensus has not been reflected in the FIGO staging system. Thus, a patient with 1 positive lymph node and a patient with 10 positive nodes are both classified as stage III tumor. Interestingly, the bilaterality of nodal involvement yielded a better prognosis if the total number of metastatic nodes was more than 7. In other words, the prognosis of a patient with 4 metastatic nodes in each groin (total = 8) seemed to be better than the prognosis of a patient with 8 metastatic nodes in 1 groin. This consideration might seem trivial but is completely ignored by the FIGO staging. The predictions of our nomogram discriminated better than did the FIGO stage (0.83 compared with 0.78, P = .01). In this regard, our nomogram represents an improvement over counseling strictly on the basis of the FIGO staging system by offering a more discriminating method of prediction. Moreover, our nomogram provides outcome prediction for cases where inguinal lymph node dissection has been omitted. One should note that our report confirms the findings of previous studies suggesting that omission of lymph node resection has an unacceptable level of fatal recurrence for patients treated for frankly invasive vulvar carcinoma.14,15 Nevertheless, a recent paper reported that elderly patients (> 80 years) had a 23.4 relative risk to have either inadequate treatments or no treatments at all.16
The nomograms provide probability estimates that might be useful at an individual level. For example, a 80-year-old patient (18 points) with a T2 tumor (10 points), 5-mm depth of invasion (43 points), and 1 metastatic node (22 points) with free margins (0 points) scores a total of 93 points, which yields an 85% probability of 2-year survival and 78% probability of 5-year survival. In a case of positive margins, the total score would be 136 (93 + 43) points, with an estimated 55% probability of 2-year survival and 40% probability of 5-year survival. Our nomogram includes more clinical and pathologic factors than the FIGO staging by adding predictive variables, such as age, bilateral node involvement, depth of invasion, and margin status. This allows the clinician to achieve a better estimation of the relapse probability of an individual patient. One should note that our nomogram also incorporates variables that are statistically insignificantly associated with survival. If the model uses only statistically significant variables, they tend to exert an inappropriately large influence, resulting in falsely narrowed confidence intervals that make the nomogram seem more accurate than it is, with poor generalizability as a consequence. For example, we kept bilaterality of nodal involvement and the interaction with the number of nodes involved in our model for 2 reasons. First, bilateral involvement is a major component of FIGO staging; therefore, we decided to use previous knowledge to build the best predictor. Second, heuristically, the effect of the total number of metastatic nodes will be modified by the unilaterality or the bilaterality of the involvement. This illustrates that the aim of the Cox model in our paper is to cover the maximal variability of cases and not to identify the independent predictors of outcome.
There are some limitations to our study that must be acknowledged. Patients were followed up for a mean of 42 months only. However, most regional and distant relapses occur in the first 2 years after the initial treatment. Our nomograms were not perfectly accurate, especially for the group of patients with high risk of relapse. Nonetheless, it merely improves on the existing ability to predict patient local and distant relapse. The reason is that most of patients have in fact a low risk of relapse, therefore the quartile of patients with a high risk of relapse is a very heterogeneous group. One should note that we did not include time period and therapeutic modalities in our nomograms. Actually, during the study period, improvements in terms of extent of surgery and morbidity have been realized, but they evidently did not translate into an improvement of survival (data not shown). On the other hand, the long observational period (24 years) is not a concern, because survival after vulvar cancer has not changed during the study period. Concerning the therapeutic modalities, their inclusion in a prognostic model might be inappropriate. First, adjuvant treatments are given because of the presence of adverse prognostic factors. Second, during the study period, adjuvant treatments were homogeneously administrated according to tumor characteristics and not in the context of a controlled trial. For example, only and almost all patients with more than 2 metastatic inguinofemoral nodes received inguinal and pelvic radiation (covariates radiation and metastatic nodes are not independent). Moreover, radiation cannot completely counteract the adverse prognostic effect of metastatic nodes; therefore, adjuvant radiation is associated with a poor outcome in a survival model. Evidently its inclusion in a nomogram is inappropriate if radiation was assigned homogeneously according to nodal status.
In conclusion, we have developed an internally and externally validated nomogram for predicting local and distant relapse of vulvar cancer at 2 and 5 years. The tool is accurate and seems to provide an improvement over existing prognostic methods for this disease.
|
Footnotes |
|---|
doi:10.1097/01.AOG.0000198639.36855.e9
|
REFERENCES |
|---|
|
TOP
ABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
2. Kattan MW, Scardino PT. Prediction of progression: nomograms of clinical utility. Clin Prostate Cancer 2002;1:90–6.[Medline]
3. International Federation of Gynecology and Obstetrics (FIGO). Changes in gynecologic cancer staging by the International Federation of Gynecology and Obstetrics. Am J Obstet Gynecol 1990;162:610–611.
4. Preti M, Ronco G, Ghiringhello B, Micheletti L. Recurrent squamous cell carcinoma of the vulva: clinicopathologic determinants identifying low risk patients. Cancer 2000;88:1869–76.[Medline]
5. Rouzier R, Haddad B, Plantier F, Dubois P, Pelisse M, Paniel BJ. Local relapse in patients treated for squamous cell vulvar carcinoma: incidence and prognostic value. Obstet Gynecol 2002;100:1159–67.
6. Rouzier R, Haddad B, Dubernard G, Dubois P, Paniel BJ. Inguinofemoral dissection for carcinoma of the vulva: effect of modifications of extent and technique on morbidity and survival. J Am Coll Surg 2003;196:442–50.[Medline]
7. Lawless JF, Singhal K. Efficient screening of nonnormal regression models. Biometrics 1978;34:318–27.
8. Durrleman S, Simon R. Flexible regression models with cubic splines. Stat Med 1989;8:551–61.[Medline]
9. Alzola C, Harrell F. An introduction to S and the Hmisc and Design Libraries. Available at: http://cran.r-project.org/doc/contrib/Alzola+Harrell-Hmisc-Design-Intro.pdf. Retrieved December 13, 2005.
10. Harrell FE Jr. Regression modeling strategies, with applications to linear models, logistic regression, and survival analysis. New York (NY): Springer; 2001.
11. Vanderbilt University. Hmisc: a package of miscellaneous S functions. Available at: http://biostat.mc.vanderbilt.edu/s/Hmisc. Retreived November 30, 2005.
12. R Development Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing. Available at: http://www.R-project.org. Retrieved November 29, 2005.
13. Hacker NF. Current management of early vulvar cancer. Ann Acad Med Singapore 1998;27:688–92.[Medline]
14. Rouzier R, Morice P, Haie-Meder C, Lhomme C, Avril MF, Duvillard P, et al. Prognostic significance of epithelial disorders adjacent to invasive vulvar carcinomas. Gynecol Oncol 2001;81:414–9.[Medline]
15. Rhodes CA, Cummins C, Shafi MI. The management of squamous cell vulval cancer: a population based retrospective study of 411 cases. Br J Obstet Gynaecol 1998;105:200–5.[Medline]
16. Vlastos AT, Usel M, Beffa V, Petignat P, Neyroud-Caspar I, Bouchardy C, et al. Treatment patterns of vulvar cancer in the elderly. Surg Oncol 2004;13:187–91.[Medline]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
