|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 25, No 35 (December 10), 2007: pp. 5669-5670 © 2007 American Society of Clinical Oncology. DOI: 10.1200/JCO.2007.14.2976
Validation of a New Prognostic Index for Advanced Epithelial Ovarian Cancer: Results From Its Application to a UK-Based CohortWellcome Trust Centre for Human Genetics, University of Oxford, Oxford; Wellcome Trust Sanger Institute, Hinxton, Cambridgeshire, United Kingdom
University of Edinburgh Cancer Research Centre, Edinburgh, United Kingdom To the Editor: Establishing the prognosis of a woman with ovarian cancer is an important part of her evaluation. Recently, Teramukai et al1 presented a prognostic index (PIEPOC) for advanced epithelial ovarian cancer, developed using the clinical and survival information of 768 women in Japan with stage III or IV disease. The prognostic index (PI) was a function of age at diagnosis, performance status (PS), histological cell type, and residual tumor size. Specifically, the index PI = 1 (if age 70 years and older) + 1 (if PS 1 or 2) + 2 (if PS 3 or 4) + 1 (if mucinous or clear cell) + 2 (if residual disease 0.1 cm or greater). This index enabled three risk groups to be defined: low (PI score 0 to 2), intermediate (3), and high (4 to 6). To ensure that this and other indices may be useful in clinical practice, both clinical and statistical validations are required using independent large retrospective and prospective cohorts of patients.2 Here, we report on our attempts to validate the Japanese prognostic index using data from the Cancer Research UK Edinburgh Ovarian Cancer Database. The Edinburgh cohort was established in 1984, and consists of more than 2,000 women diagnosed with epithelial ovarian carcinoma (see Clark et al3 for a description of the cohort). To validate the transportability of PIEPOC, we consider the subset of Edinburgh patients (n = 894) diagnosed with advanced primary epithelial ovarian carcinoma between January 1984 and December 2004, who had undergone surgical debulking and were treated with first-line chemotherapy. Follow-up data were available until the end of December 2006, and the median follow-up of survivors is 4.6 years. The patient characteristics for this subset (n = 894) are presented in Table 1.
Application of PIEPOC to the Edinburgh cohort with advanced disease resulted in the following percentage of patients in each of the PI scores: 0, 4.4%; 1, 6.6%; 2, 21.3%; 3, 43.1%; 4, 21.0%; and 5, 3.7%. These translate into risk groups: low or PI scores 0 to 2 (32.2%); intermediate or PI = 3 (43.1%); and high or PI score more than 3 (24.7%). Figure 1 shows the survival curves for these groups, and though there is strong evidence of global differences (log-rank P < .001), there is overlap or less discrimination between the intermediate- and high-risk groups at 5 years after diagnosis. The 2-year survival probabilities (95% CI) for the low- to high-risk groups are 0.651 (0.593 to 0.703), 0.373 (0.325 to 0.421), and 0.250 (0.195 to 0.309), respectively. The 5-year survival probabilities (95% CI) for the low- to high-risk groups are 0.287 (0.233 to 0.243), 0.109 (0.079 to 0.145), and 0.063 (0.034 to 0.104), respectively. The c-index—a measure of the ability to distinguish between high- and low-risk patients (discrimination)—for the model was 0.625, which indicates that predictions are better than random (> 0.5). The calibration of predicted probabilities (at 5 years) with observed probabilities using an estimate of slope shrinkage4 was 0.996. Both these measures of predictive performance are similar to those reported by Teramukai et al.1 We found that recategorizing the low- to high-risk groups using PI scores 0 to 1, 2 to 3, and more than 3, respectively, led to greater discrimination between the survival curves (plots not shown), with 5-year survival probabilities (95% CI) for the low- to high-risk groups being 0.374 (0.273 to 0.474), 0.154 (0.124 to 0.186), and 0.063 (0.034 to 0.104), respectively. When fitting a multivariate model with all prognostic factors in the PI, only age (  69 vs > 69 years) was found to be statistically nonsignificant (P > .1). The inclusion of ascites and grade in the model did not significantly improve the model performance (log-likelihood ratio test P value > .17).
In summary, we confirm that PIEPOC is predictive in a second large data set, and that it is transportable to a Western European population. We also show that in this population, slight recategorization of the low- to high-risk groups may lead to greater discrimination between the survival curves. One potential difficulty with PIEPOC is its use of performance status. The predictive value of performance status has been confirmed in multiple studies, but its assessment can vary between observers and according to the time of evaluation in relation to debulking surgery. The strengths of the index are its simplicity and ease of application in clinical practice. This has largely been made possible by removal of the early disease component, which is considered by some to have a different etiology.5 One wonders if developing separate indices for clear-cell and mucinous histologies would allow further refinement in prognostic evaluation. AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST The author(s) indicated no potential conflicts of interest.
REFERENCES
1. Teramukai S, Ochiai K, Tada H, et al: PIEPOC: A new prognostic index for advanced epithelial ovarian cancer: Japan Multinational Trial Organization OC01-01. J Clin Oncol 25:3302-3306, 2007 2. Altman DG, Royston P: What do we mean by validating a prognostic model? Stat Med 19:453-473, 2000[CrossRef][Medline] 3. Clark TG, Stewart ME, Altman DG, et al: A prognostic model for ovarian cancer. Br J Cancer 85:944-952, 2001[CrossRef][Medline] 4. Harrell FE Jr, Lee KL, Mark DB: Multivariable prognostic models: Issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 15:361-387, 1996[CrossRef][Medline] 5. Hogg R, Friedlander M: Biology of epithelial ovarian cancer: Implications for screening women at high genetic risk. J Clin Oncol 22:1315-1327, 2004 Related Reply
Related Article
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2007 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
