Serge Koscielny

Prediction of cancer outcome with microarrays: a multiple random validation strategy.

BACKGROUND General studies of microarray gene-expression profiling have been undertaken to predict cancer outcome. Knowledge of this gene-expression profile or molecular signature should improve treatment of patients by allowing treatment to be tailored to the severity of the disease. We reanalysed data from the seven largest published studies that have attempted to predict prognosis of cancer patients on the basis of DNA microarray analysis. METHODS The standard strategy is to identify a molecular signature (ie, the subset of genes most differentially expressed in patients with different outcomes) in a training set of patients and to estimate the proportion of misclassifications with this signature on an independent validation set of patients. We expanded this strategy (based on unique training and validation sets) by using multiple random sets, to study the stability of the molecular signature and the proportion of misclassifications. FINDINGS The list of genes identified as predictors of prognosis was highly unstable; molecular signatures strongly depended on the selection of patients in the training sets. For all but one study, the proportion misclassified decreased as the number of patients in the training set increased. Because of inadequate validation, our chosen studies published overoptimistic results compared with those from our own analyses. Five of the seven studies did not classify patients better than chance. INTERPRETATION The prognostic value of published microarray results in cancer studies should be considered with caution. We advocate the use of validation by repeated random sampling.

False discovery rate, sensitivity and sample size for microarray studies

MOTIVATION In microarray data studies most researchers are keenly aware of the potentially high rate of false positives and the need to control it. One key statistical shift is the move away from the well-known P-value to false discovery rate (FDR). Less discussion perhaps has been spent on the sensitivity or the associated false negative rate (FNR). The purpose of this paper is to explain in simple ways why the shift from P-value to FDR for statistical assessment of microarray data is necessary, to elucidate the determining factors of FDR and, for a two-sample comparative study, to discuss its control via sample size at the design stage. RESULTS We use a mixture model, involving differentially expressed (DE) and non-DE genes, that captures the most common problem of finding DE genes. Factors determining FDR are (1) the proportion of truly differentially expressed genes, (2) the distribution of the true differences, (3) measurement variability and (4) sample size. Many current small microarray studies are plagued with large FDR, but controlling FDR alone can lead to unacceptably large FNR. In evaluating a design of a microarray study, sensitivity or FNR curves should be computed routinely together with FDR curves. Under certain assumptions, the FDR and FNR curves coincide, thus simplifying the choice of sample size for controlling the FDR and FNR jointly.

Breast cancer: Relationship between the size of the primary tumour and the probability of metastatic dissemination

The relationship between the size of the primary tumour upon initial treatment and the incidence of distant metastasis during the course of the disease was investigated using data from 2648 breast cancers treated at the Institut Gustave Roussy between 1954 and 1972. This analysis suggests the existence for each tumour of a critical volume (threshold) at which the first remote metastasis is initiated. The correlation between the size of the primary tumour and the probability of metastatic dissemination was assessed as well as the influence on this correlation of two prognostic indicators: histological grade and number of involved lymph nodes. It was found that the threshold volume is strongly correlated with the number of involved lymph nodes and the histological grading.

Advanced Hepatocellular Carcinoma: Early Evaluation of Response to Bevacizumab Therapy at Dynamic Contrast-enhanced US with Quantification—Preliminary Results

PURPOSE To investigate whether there is any correlation between standard efficacy endpoints-specifically, tumor response, progression-free survival, and overall survival-and tumor perfusion parameters measured by using dynamic contrast material-enhanced ultrasonography (US) in patients with advanced hepatocellular carcinoma (HCC) treated with bevacizumab. MATERIALS AND METHODS The institutional review board approved the study, and all patients provided written informed consent before their enrollment. Between June 3, 2005, and September 28, 2007, 42 patients (33 men, nine women; median age, 62 years; age range, 23-84 years) participated in this phase II study of single-agent bevacizumab treatment. Tumor response (based on RECIST [response evaluation criteria in solid tumors]) at 2 months was assessed in 37 patients, and progression-free survival and overall survival were assessed in all 42 patients. Dynamic contrast-enhanced US (ie, dynamic US) was performed before treatment (day 0); on days 3, 7, 14, and 60 after treatment; and every 2 months thereafter. Tumor perfusion parameters were estimated quantitatively from contrast material uptake curves constructed from raw linear data. The changes in dynamic US functional parameters between day 0 and the later time points were compared between treatment responders and nonresponders by using nonparametric tests. Given multiple comparisons, P < .001 indicated significance. RESULTS The percentage decrease in several dynamic US parameters between day 0 and day 3 showed trends toward correlation with (a) tumor response in terms of total area under the time-intensity curve (AUC) (P = .02), AUC during wash in (P = .04), AUC during washout (P = .02), and time to peak intensity (P = .03); (b) progression-free survival in terms of time to peak intensity (P = .028); and (c) overall survival in terms of AUC (P = .002) and AUC during washout (P = .003). CONCLUSION Dynamic US can be used to quantify dynamic changes in tumor vascularity as early as 3 days after bevacizumab administration in patients with HCC. These early changes in tumor perfusion may be predictive of tumor response at 2 months, progression-free survival, and overall survival, and they may be potential surrogate measures of the effectiveness of antiangiogenic therapy in patients with HCC.

Metastatic Renal Cell Carcinoma Treated with Sunitinib: Early Evaluation of Treatment Response Using Dynamic Contrast-Enhanced Ultrasonography

Purpose: To determine the utility of dynamic contrast-enhanced ultrasonography (DCE-US) as a prognostic tool for metastatic renal cell carcinoma patients receiving sunitinib and to identify DCE-US parameters that correlate with early treatment response. Experimental Design: Thirty-eight patients received 50 mg/d sunitinib on schedule 4/2 (4 weeks on followed by 2 weeks off treatment). After two cycles, response evaluation criteria in solid tumors were used to classify patients as responders or nonresponders. DCE-US evaluations were done before treatment and at day 15; variations between days 0 and 15 were calculated for seven DCE-US functional parameters and were compared for responders and nonresponders. The correlation between DCE-US parameters and disease-free survival (DFS) and overall survival (OS) was assessed. Results: The ratio between DCE-US examinations at baseline and day 15 significantly correlated with response in five of the seven DCE-US parameters. Two DCE-US parameters (time to peak intensity and slope of the wash-in) were significantly associated with DFS; time to peak intensity was also significantly associated with OS. Conclusions: DCE-US is a useful tool for predicting the early efficacy of sunitinib in metastatic renal cell carcinoma patients. Robust correlations were observed between functional parameters and classic assessments, including DFS and OS. Clin Cancer Res; 16(4); 1216–25

Full dose reirradiation combined with chemotherapy after salvage surgery in head and neck carcinoma

The purpose of this study was to analyze the tolerance and efficacy of full dose reirradiation combined with chemotherapy in patients with head and neck carcinoma (HNC) with a high risk of recurrence after salvage surgery.

Evaluation of contrast-enhanced color Doppler ultrasound for the quantification of angiogenesis in vivo.

Lassau N, Koscielny S, Opolon P, et al. Evaluation of contrast-enhanced color Doppler ultrasound for the quantification of angiogenesis in vivo. Invest Radiol 2001;36:50–55. rationale and objectives.To evaluate contrast-enhanced color Doppler ultrasonography (CDUS) aimed at quantifying angiogenesis in vivo. methods.Eight colon tumors xenografted into mice were studied with CDUS both without and after injection of SHU 508A. The number of intratumor vessels and pedicles visualized with CDUS and the value of the acoustic intensity quantified by computer were compared with the evaluation of angiogenesis by study of digitized histological slides. results.SHU 508A increased the number of intratumor vessels and pedicles depicted by CDUS, which could also be quantified by acoustic intensity. The comparison of CDUS and digitized histological slide studies showed that the detection threshold after SHU 508A administration allowed visualization of small vessels (40 &mgr;m in diameter). The “hot spots” seen on histology were superimposable onto vessels seen with CDUS. Regions in which no vessels were detected by CDUS corresponded to regions in which vascularization was absent by histological analysis. conclusion.Injection of SHU 508A significantly increased vessel detection with CDUS.

Busulfan disposition and hepatic veno-occlusive disease in children undergoing bone marrow transplantation

Hepatic veno-occlusive disease (HVOD) is a frequent life-threatening toxicity in patients undergoing bone marrow transplantation (BMT) after the administration of a high-dose busulfan-containing regimen. Recent studies have shown that the morbidity and mortality of HVOD may be reduced in adults by pharmacologically guided dose adjustment of busulfan. We analyzed the pharmacodynamic relationship between busulfan disposition and HVOD in 61 children (median age, 5.9 years) with malignant disease. Busulfan, given at a dose ranging from 16 mg/kg to 600 mg/m2, was combined with one or two other alkylating agents (cyclophosphamide, melphalan, thiotepa). Only 3 patients received the standard busulfan/cyclophosphamide (BUCY) regimen. A total of 24 patients (40%) developed HVOD, which resolved in all cases. A pharmacokinetics study confirmed the previously reported wide interpatient variability in busulfan disposition but did not reveal any significant alteration in children with HVOD. The mean area under the concentration-time curve (AUC) after the first dose of busulfan was higher in patients with HVOD (6,811±2,943 ng h ml−1) than in patients without HVOD (5,760±1,891 ng h ml−1;P=0.10). This difference reflects the higher dose of busulfan given to patients with HVOD. No toxic level could be defined and, moreover, none of the toxic levels identified in adults were relevant. The high incidence of HVOD in children given 600 mg/m2 busulfan may be linked to the use of more intensive than usual high-dose chemotherapy regimens and/or drug interactions. Before the prospective evaluation of busulfan dose adjustment in children, further studies are required to demonstrate firmly the existence of a pharmacodynamic relationship in terms of toxicity and allogeneic engraftment, especially when busulfan is combined with cyclophosphamide. The maximal tolerated and minimal effective AUCs in children undergoing BMT are likely to depend mainly upon the disease, the nature of the combined high-dose regimen, and the type of bone marrow transplant.

Natural history of human breast cancer: Recent data and clinical implications

SummaryThis study of the natural history of human breast cancer was based on the analysis of a series of 3000 patients treated by radical mastectomy at a single institution (Institut Gustave Roussy) at a time when adjuvant chemotherapy was not prescribed. The follow-up of the patients ranged from 15 to 30 years; for each patient the tumor size, the number of involved axillary nodes, and the histological grade were prospectively registered.A highly significant correlation was found between tumor size and the probability of distant metastatic dissemination. The distribution of tumor sizes at metastatic spread was log-normal with a median diameter equal to 3.5 cm.The patients were subdivided into 3 groups according to the histological grade. In each subgroup there was a significant correlation between tumor size and the probability of distant spread; the distributions were log-normal and the median size was markedly larger for grade 1 tumors. Moreover the proportion of grade 1 tumors was higher in small tumors than in large ones while the reverse was observed for grade 3 tumors; these data suggest that during their growth tumors progress towards higher grades.One of the chief fundamental characteristics of a tumor seems to be its propensity for axillary node invasion. The orderly pattern of nodal involvement makes it possible to calculate the tumor size at invasion of the first axillary node in each subset of patients. A strong and highly significant correlation exists between the size of the tumor at initation of distant metastasis and at invasion of the first lymph node. However the capacity for lymphatic spread is, on average, acquired much earlier than the capacity for metastatic spread.With a simple model based on these data it was possible to compute the proportion of patients with occult metastases as a function of tumor size, histological grade, and number of involved axillary nodes.Early invasion of axillary nodes is associated with a rapid growth rate of the primary tumor (or a high S-phase fraction). However each of these variables has an independent prognostic significance; the S-phase fraction appears as one of the strongest prognostic indicators.A model of tumor growth was used to assess the impact of screening procedures on the proportion of patients with distant metastases. The predictions of the model are consistent with the results of the screening programs. The model was used to predict the influence of the interval between mammographies on the proportion of patients with distant metastases and on the size of these metastases. Since the curability of distant metastases by adjuvant chemotherapy is, to a large extent, governed by their size, the model can predict the cumulative effect of screening procedures and adjuvant chemotherapy on the proportion of patients with overt distant metastases.

Interpretation of microarray data in cancer.

Microarray studies aim at identifying homogeneous subtypes of cancer patients, searching for differentially expressed genes in tumours with different characteristics, or predicting the prognosis of patients. Using breast cancer as an example, we discuss the hypotheses underlying these studies, their power, and the validity and the clinical usefulness of the findings.