CA-125 AUC as a new prognostic factor for patients with ovarian cancer

Prediction of survival outcomes in patients with epithelial ovarian cancer using machine learning methods

OBJECTIVES

The aim of this study was to develop a new prognostic classification for epithelial ovarian cancer (EOC) patients using gradient boosting (GB) and to compare the accuracy of the prognostic model with the conventional statistical method.

METHODS

Information of EOC patients from Samsung Medical Center (training cohort, n=1,128) was analyzed to optimize the prognostic model using GB. The performance of the final model was externally validated with patient information from Asan Medical Center (validation cohort, n=229). The area under the curve (AUC) by the GB model was compared to that of the conventional Cox proportional hazard regression analysis (CoxPHR) model.

RESULTS

In the training cohort, the AUC of the GB model for predicting second year overall survival (OS), with the highest target value, was 0.830 (95% confidence interval [CI]=0.802-0.853). In the validation cohort, the GB model also showed high AUC of 0.843 (95% CI=0.833-0.853). In comparison, the conventional CoxPHR method showed lower AUC (0.668 (95% CI=0.617-0.719) for the training cohort and 0.597 (95% CI=0.474-0.719) for the validation cohort) compared to GB. New classification according to survival probability scores of the GB model identified four distinct prognostic subgroups that showed more discriminately classified prediction than the International Federation of Gynecology and Obstetrics staging system.

CONCLUSION

Our novel GB-guided classification accurately identified the prognostic subgroups of patients with EOC and showed higher accuracy than the conventional method. This approach would be useful for accurate estimation of individual outcomes of EOC patients.

Integrated multiomic predictors for ovarian cancer survival

Increasingly affordable high-throughput molecular profiling technologies have made feasible the measurement of omics-wide interindividual variations for the purposes of predicting cancer prognosis. While multiple types of genetic, epigenetic and expression changes have been implicated in ovarian cancer, existing prognostic biomarker strategies are constrained to analyzing a single class of molecular variations. The extra predictive power afforded by the integration of multiple omics types remains largely unexplored. In this study, we performed integrative analysis on tumor-based exome-, transcriptome- and methylome-wide molecular profiles from The Cancer Genome Atlas (TCGA) for variations in cancer-relevant genes to construct robust, cross-validated multiomic predictors for ovarian cancer survival. These integrated polygenic survival scores (PSSs) were able to predict 5-year overall (OS) and progression-free survival in the Caucasian subsample with high accuracy (AUROC = 0.87 and 0.81, respectively). These findings suggest that the PSSs are able to predict long-term OS in TCGA patients with accuracy beyond that of previously proposed protein-based biomarker strategies. Our findings reveal the promise of an integrated omics-based approach in enhancing existing prognostic strategies. Future investigations should be aimed toward prospective external validation, strategies for standardizing application and the integration of germline variants.

Expression of miR-146a in patients with ovarian cancer and its clinical significance

The aim of the present retrospective study was to compare microRNA (miR)-146a expression levels in primary tumors and omental metastases of 48 patients, who had undergone surgery for advanced ovarian serous cancer. Possible correlations between miR-146a expression level and clinicopathological features were investigated, including chemosensitivity and survival. miR-146a was evaluated in formalin-fixed, paraffin-embedded samples. miR-146a expression level in primary tumors was demonstrated to be increased in comparison with normal ovary tissues (P=0.02) and metastases (P=0.01). A negative correlation was demonstrated between miR-146a expression in primary tumors and serum levels of cancer antigen 125 (R=-0.37; P=0.03) and Risk of Malignancy Algorithm index (R=-0.79; P=0.0007). Overall survival positively correlated with miR-146a expression in primary tumor tissue samples (R=0.38; P=0.01). Probability of survival was decreased in patients with low miR-146a expression levels in primary tumor tissues (hazard ratio=0.21; P=0.003). Lower levels of miR-146a in primary tumor tissue samples were correlated with a shorter progression-free survival (P=0.04) and platinum-resistance of metastases (P=0.006). In conclusion, miR-146a may be a prognostic marker for serous ovarian cancer.

Intestinal Microbiota and Relapse After Hematopoietic-Cell Transplantation

Purpose The major causes of mortality after allogeneic hematopoietic-cell transplantation (allo-HCT) are relapse, graft-versus-host disease (GVHD), and infection. We have reported previously that alterations in the intestinal flora are associated with GVHD, bacteremia, and reduced overall survival after allo-HCT. Because intestinal bacteria are potent modulators of systemic immune responses, including antitumor effects, we hypothesized that components of the intestinal flora could be associated with relapse after allo-HCT. Methods The intestinal microbiota of 541 patients admitted for allo-HCT was profiled by means of 16S ribosomal sequencing of prospectively collected stool samples. We examined the relationship between abundance of microbiota species or groups of related species and relapse/progression of disease during 2 years of follow-up time after allo-HCT by using cause-specific proportional hazards in a retrospective discovery-validation cohort study. Results Higher abundance of a bacterial group composed mostly of Eubacterium limosum in the validation set was associated with a decreased risk of relapse/progression of disease (hazard ratio [HR], 0.82 per 10-fold increase in abundance; 95% CI, 0.71 to 0.95; P = .009). When the patients were categorized according to presence or absence of this bacterial group, presence also was associated with less relapse/progression of disease (HR, 0.52; 95% CI, 0.31 to 0.87; P = .01). The 2-year cumulative incidences of relapse/progression among patients with and without this group of bacteria were 19.8% and 33.8%, respectively. These associations remained significant in multivariable models and were strongest among recipients of T-cell-replete allografts. Conclusion We found associations between the abundance of a group of bacteria in the intestinal flora and relapse/progression of disease after allo-HCT. These might serve as potential biomarkers or therapeutic targets to prevent relapse and improve survival after allo-HCT.

Clinical Significance of CA125 Level after the First Cycle of Chemotherapy on Survival of Patients with Advanced Ovarian Cancer

PURPOSE

To determine the most powerful cancer antigen 125 (CA125)-related prognostic factor for advanced epithelial ovarian cancer (EOC) and to identify cut-off values that distinguish patients with a poor prognosis from those with a good prognosis.

MATERIALS AND METHODS

We included 223 patients who received staging laparotomy and were diagnosed with stage IIC-IV serous EOC. Cox regression analysis was used to determine the most significant prognostic factor among the following variables: serum CA125 before surgery and after the first, second, and sixth cycles of chemotherapy; the nadir CA125 value; the relative percentage change in CA125 levels after the first and second cycles of chemotherapy compared to baseline CA125; CA125 half-life; time to nadir; and time to normalization of the CA125 level.

RESULTS

The CA125 level after the first chemotherapy cycle was the most significant independent prognostic factor for overall survival (OS). Time to normalization (p=0.028) and relative percentage change between CA125 levels at baseline and after the first chemotherapy cycle (p=0.021) were additional independent prognostic factors in terms of OS. The CA125 level after the first chemotherapy cycle (p=0.001) and time to normalization (p<0.001) were identified as independent prognostic factors for progression free survival (PFS).

CONCLUSION

Among well-established CA125-related prognostic factors, serum CA125 levels after the first cycle of chemotherapy and time to normalization were the most significant prognostic factors for both OS and PFS.

A joint model based on longitudinal CA125 in ovarian cancer to predict recurrence

AIMS

To develop a new package of joint model to fit longitudinal CA125 in epithelial ovarian cancer relapse.

PATIENTS & METHODS

Included were 305 epithelial ovarian cancer patients who reached complete remission after cytoreductive surgery and first-line chemotherapy. Univariate and multivariate analysis with a joint model was performed to select independent risk factors, which were subsequently combined to predict recurrence.

RESULTS

Independent factors were longitudinal CA125, age, stage and residual tumor size (p < 0.05). Prediction of recurrence with these factors had an average of 80.7% accuracy, 5.6-10.7% better than kinetic factors.

CONCLUSION

The new package of joint model fits longitudinal CA125 well. Potential application can be extended to other biomarkers.

CA125-related tumor cell kinetics variables after chemotherapy in advanced ovarian cancer: a systematic review

Various kinetic parameters, based on a minimum of two time points, have been built with CA125 determinations. The aim of this study is to review studies about the clinical application of CA125-related tumor cell kinetics variables in patients with advanced ovarian cancer (AOC) receiving chemotherapy. A literature search for studies about CA125-related variables in patients with AOC was undertaken on three databases, by predefined search criteria, and a selection of studies was performed. Sixty-two studies were selected. CA125-related variables were summarized in three groups: response-related, time-to-event, and other CA125-related tumor cell kinetics variables. Even though CA125 changes and half-life after chemotherapy were the most studied, other variables and two models have been well defined, and often showed an interesting power to predict survival. These kinetics variables are related to the CA125 regression curve, pre- and post-chemotherapy kinetics, or are variables inferred from a population model of CA125 kinetics.

Comparison of LUMIPULSE(®) G1200 With Kryptor and Modular E170 for the Measurement of Seven Tumor Markers

BACKGROUND

Tumor marker measurements are becoming essential for prognosis and follow-up of patients in oncology. In this context, we aimed to compare a new analyzer, Lumipulse(®) G1200 (Fujirebio group, distributed in Europe by the Innogenetics group) with Kryptor(®) (Thermo Fisher Scientific B.R.A.H.M.S, Asnières, France) and Modular(®) Elecsys E170 (Roche Diagnostics, Meylan, France) for the measurement of seven tumor markers: PSA, AFP, CEA, CA 15-3, CA 125, CA 19-9, and Cyfra 21-1.

METHODS

A total of 471 serum samples from patients with elevated tumor markers and 100 serum from healthy patients were analyzed with Lumipulse(®) G1200 and either Kryptor(®) (for AFP) or Modular(®) (for the six other markers).

RESULTS

The good precision of Lumipulse(®) G1200 assays was confirmed with CVs < 2.5% and < 5.0%, obtained, respectively, for within-run imprecision and intermediate imprecision (except for Cyfra 21-1: CV < 13%). For all markers, Lumipulse results were well correlated with Modular or Kryptor results (r ≥ 0.94). Concordance of results interpretation was > 95% and tumor marker kinetics were all similar.

CONCLUSION

We confirmed the analytical performances of Lumipulse(®) tumor marker assays except for the CYFRA 21-1 assay for which performances were poor in this study. We noticed a few discrepancies for the CEA assay. Besides, values obtained for CA 19-9 were higher with Lumipulse leading to a bias (slope = 1.5). But for the four other tumor markers assays (PSA, AFP, CA 125, CA 15-3), the results were directly transferable between Lumipulse and Kryptor or Modular, thus facilitating an eventual substitution of one system by another.

A critical review of the analytical approaches for circulating tumor biomarker kinetics during treatment

Changes in serum tumor biomarkers may indicate treatment efficacy. Traditional tumor markers may soon be replaced by novel serum biomarkers, such as circulating tumor cells (CTCs) or circulating tumor nucleic acids. Given their promising predictive values, studies of their kinetics are warranted. Many methodologies meant to assess kinetics of traditional marker kinetics during anticancer treatment have been reported. Here, we review the methodologies, the advantages and the limitations of the analytical approaches reported in the literature. Strategies based on a single time point were first used (baseline value, normalization, nadir, threshold at a time t), followed by approaches based on two or more time points [half-life (HL), percentage decrease, time-to-events…]. Heterogeneities in methodologies and lack of consideration of inter- and intra-individual variability may account for the inconsistencies and the poor utility in routine. More recently, strategies based on a population kinetics approach and mathematical modeling have been reported. The identification of equations describing individual kinetic profiles of biomarkers may be an alternative strategy despite its complexity and higher number of necessary measurements. Validation studies are required. Efforts should be made to standardize biomarker kinetic analysis methodologies to ensure the optimized development of novel serum biomarkers and avoid the pitfalls of traditional markers.

The strong prognostic value of KELIM, a model-based parameter from CA 125 kinetics in ovarian cancer: data from CALYPSO trial (a GINECO-GCIG study)

BACKGROUND

Unexpected results were recently reported about the poor surrogacy of Gynecologic Cancer Intergroup (GCIG) defined CA-125 response in recurrent ovarian cancer (ROC) patients. Mathematical modeling may help describe CA-125 decline dynamically and discriminate prognostic kinetic parameters.

METHODS

Data from CALYPSO phase III trial comparing 2 carboplatin-based regimens in ROC patients were analyzed. Based on population kinetic approach, serum [CA-125] concentration-time profiles during first 50 treatment days were fit to a semi-mechanistic model with following parameters: "d[CA-125]/dt=(KPROD∗exp (BETA∗t))∗Effect-KELIM∗[CA-125]" with time, t; tumor growth rate, BETA; CA-125 tumor production rate, KPROD; CA-125 elimination rate, KELIM and K-dependent treatment indirect Effect. The predictive values of kinetic parameters were tested regarding progression-free survival (PFS) against other reported prognostic factors.

RESULTS

Individual CA-125 kinetic profiles from 895 patients were modeled. Three kinetic parameters categorized by medians had predictive values using univariate analyses: K; KPROD and KELIM (all P<0.001). Using Cox multivariate analysis, 5 independent predictors of PFS remained significant: GCIG CA-125 response (favoring carboplatin-paclitaxel arm), treatment arm, platinum free-interval, measurable lesions and KELIM (HR=0.53; 95% CI 0.45-0.61; P<0.001).

CONCLUSIONS

Mathematical modeling of CA-125 kinetics in ROC patients enables understanding of the time-change components during chemotherapy. The contradictory surrogacy of GCIG-defined CA-125 response was confirmed. The modeled CA-125 elimination rate KELIM, potentially assessable in routine, may have promising predictive value regarding PFS. Further validation of this predictive marker is warranted.

Longitudinal monitoring of CA125 levels provides additional information about survival in ovarian cancer

Background

We investigated the prognostic impact of changes in serum CA125 levels during the first 3 months of therapy in ovarian cancer.

Methods

A case series of 170 ovarian cancer patients treated at Cancer Treatment Centers of America. Based on CA125 levels at baseline and 3 months, patients were classified into 4 groups: 1) Normal (0-35 U/ml) at baseline and three months; 2) High (>35 U/ml) at baseline, normal at three months; 3) Normal at baseline, high at 3 months; 4) High at baseline and three months. Kaplan Meier method was used to calculate survival across the 4 categories.

Results

Of 170 patients, 36 were newly diagnosed while 134 had received prior treatment. 25 had stage I disease at diagnosis, 15 stage II, 106 stage III and 14 stage IV. The median age at presentation was 54.2 years (range 23.1 - 82.5 years). At baseline, 31 patients had normal (0-35 U/ml) serum CA125 levels while 139 had high (>35 U/ml) levels. At 3 months, 59 had normal while 111 had high levels. Patients with a reduced CA125 at 3 months had a significantly better survival than those with increased CA125 at 3 months. Patients with normal values of CA125 at both baseline and 3 months had the best overall survival.

Conclusions

These data show that reduction in CA125 after 3 months of therapy is associated with better overall survival in ovarian cancer. Patients without a significant decline in CA125 after 3 months of therapy have a particularly poor prognosis.

Role of CA125 in predicting ovarian cancer survival - a review of the epidemiological literature

CA125 is the gold standard tumor marker in ovarian cancer. Serum level of CA125 is used to monitor response to chemotherapy, relapse, and disease progression in ovarian cancer patients. Thus, it is reasonable to investigate whether CA125 may have utility as a prognostic indicator as well in ovarian cancer. A large number of epidemiological studies have been carried out to this effect. This review summarizes all available epidemiological literature on the association between CA125 levels and survival in ovarian cancer. To place these studies in context, we provide some background information on CA125 and its role in ovarian cancer.

Do serum biomarkers really measure breast cancer?

Background

Because screening mammography for breast cancer is less effective for premenopausal women, we investigated the feasibility of a diagnostic blood test using serum proteins.

Methods

This study used a set of 98 serum proteins and chose diagnostically relevant subsets via various feature-selection techniques. Because of significant noise in the data set, we applied iterated Bayesian model averaging to account for model selection uncertainty and to improve generalization performance. We assessed generalization performance using leave-one-out cross-validation (LOOCV) and receiver operating characteristic (ROC) curve analysis.

Results

The classifiers were able to distinguish normal tissue from breast cancer with a classification performance of AUC = 0.82 ± 0.04 with the proteins MIF, MMP-9, and MPO. The classifiers distinguished normal tissue from benign lesions similarly at AUC = 0.80 ± 0.05. However, the serum proteins of benign and malignant lesions were indistinguishable (AUC = 0.55 ± 0.06). The classification tasks of normal vs. cancer and normal vs. benign selected the same top feature: MIF, which suggests that the biomarkers indicated inflammatory response rather than cancer.

Conclusion

Overall, the selected serum proteins showed moderate ability for detecting lesions. However, they are probably more indicative of secondary effects such as inflammation rather than specific for malignancy.