When is a prognostic factor useful? A guide for the perplexed

Traditionally, a number of variables have been used to predict outcome in patients with early-stage breast cancer. These tests are simple to perform and relatively inexpensive. Recently, a number of new factors, eg, tumor proliferative index, nuclear DNA content, and amplification or overexpression of growth-promoting genes or oncogenes have been identified as potential predictors of outcome in patients with breast cancer. There is now increasing pressure to introduce such tests into routine clinical practice. How does a clinical practitioner identify which test, or group of tests, best predicts adverse outcome and whether any more clinically useful information is provided than with the use of more traditional factors alone? The aim of a prognostic test in breast cancer is to predict which patients are destined to develop a recurrence of cancer and those who are not. The prognostic usefulness of a test can be expressed in terms of relative risk (RR), which is the ratio of the risk of breast cancer recurrence in patients who test positive to the risk in those who test negative. Methodologic guidelines that should be satisfied by a study evaluating the predictive ability of a test include the following: (1) Was an inception cohort assembled? (2) Was the referral pattern described? (3) Were laboratory and clinical outcomes assessed in a blinded fashion? (4) Was complete follow-up achieved? (5) Was adjustment for extraneous prognostic factors carried out? (6) Were appropriate statistical methods used? An approach is suggested to help the clinician choose the test, or combination of tests, likely to discriminate between "high-" and "low-risk" patients in his/her own practice. The decision regarding what particular threshold value (risk) defined by a prognostic test (or series of tests) warrants adjuvant therapy for an individual patient is a complex one but should be based on a clear presentation of the risks and benefits to the patient.     

How relevant is hormone receptor status in the context of outcome to HER2-positive breast cancer?

Clinical outcome of patients with breast cancer is based on patient and tumor-related factors. The relevant tumor-related factors include anatomical extent and biology. Of the prognostic and predictive biological markers available, hormone receptors (defined as estrogen and progesterone receptors) and HER2 receptors, have been independently validated. Pertinent questions to be addressed include their combined impact on prognosis, their relevance in terms of sites of metastases, and whether they change in primary versus recurrent tumors. Although these questions are being addressed in clinical trials, epidemiological results, such as those derived from the National Comprehensive Cancer Network dataset, add perspective to our understanding of these two most relevant biological prognostic/predictive markers.     

Prognostic factors: Rationale and methods of analysis and integration

Summary With the proliferation of potential prognostic factors for breast cancer, it is becoming increasingly more difficult for physicians and patients to integrate the information provided by these factors into a single accurate prediction of clinical outcome. Here we review Cox's proportional hazards model, recursive partitioning, correspondence analysis, and neural networks for their respective capabilities in analyzing censored survival data in the presence of multiple prognostic factors, and we present some clinical applications where these models have been used.     

Modeling the overall survival of patients with advanced‐stage non‐small cell lung cancer using data of routine laboratory tests

Cancer patients undergo routine clinical monitoring with an array of blood tests that may carry long‐term prognostic information. We aimed to develop a new prognostic model predicting survival for patients with advanced non‐small cell lung cancer (NSCLC), based on laboratory tests commonly performed in clinical practice. A cohort of 1,161 stage IIIB or IV NSCLC patients was divided into training (n = 773) and testing (n = 388) cohorts. We analyzed the associations of 32 commonly tested laboratory variables with patient survival in the training cohort. We developed a model based on those significant laboratory variables, together with important clinical variables. The model was then evaluated in the testing cohort. Five variables, including albumin, total protein, alkaline phosphatase, blood urea nitrogen and international normalized ratio, were significantly associated with patient survival after stepwise selection. A model incorporating these variables classified patients into low‐, medium‐ and high‐risk groups with median survival of 16.9, 7.2 and 2.1 months, respectively (p < 0.0001). Compared with low‐risk group, patients in the medium‐ and high‐risk groups had a significantly higher risk of death at 1 year, with hazard ratio (HR) of 1.95 (95% CI 1.62–2.36) and 5.22 (4.30–6.34), respectively. These results were validated in the testing cohort. Overall, we developed a prognostic model relying entirely on readily available variables, with similar predictive power to those which depend on more specialized and expensive molecular assays. Further study is necessary to validate and further refine this model, and compare its performance to models based on more specialized and expensive testing.     

Validation of breast cancer prognostic models

Besides classic prognostic factors in breast cancer, there are numerous variables without fully established prognostic value. The use of all these factors in daily clinical practice may prove difficult. In this study we investigated the validity of certain reported prognostic models in our own population of patients. In a retrospective series of 244 breast cancer patients, we determined certain clinical, morphological and immunohistochemistry variables. Validity analysis of the prognostic models reported by other authors was performed by the use of their corresponding Cox equations calculated from the data published. The validation made with the Paik model without prognostic value has obtained statistical significance in our series, and variables predicting the evolution of their patients was not confirmed. The application of the Lovekin prognostic model did appear to be valid in our group of patients. It has not been possible to verify as prognostic in our patients all the models proposed by other authors.     

Prognostic value of quantitative pathologic features and DNA content in individual patients with stage I endometrial adenocarcinoma

The prognostic value of clinical features, qualitative and quantitative pathologic characteristics and steroid receptor incidence was evaluated in 50 patients with Stage I endometrial adenocarcinoma. It turned out that many of these features were prognostically important. Estrogen receptor content was not significantly associated with prognosis in our material, but patients with progesterone receptor positive lesions had a better survival than those in which the tumors were progesterone receptor negative. Multivariate analysis also clarified that only three features in combination had independent significance: mean shortest nuclear axis, DNA index, and depth of myometrial invasion (in sequence of decreasing importance). The prognostic rule consisting of these features overshadowed the value of all other features investigated. The overriding prognostic value of this highly reproducible rule was clear from the complete separation of 27 survivors and six nonsurvivors in the learning set of 33 patients. In an independent test set, all three nonsurvivors and 13 of the 14 survivors were correctly classified, thus confirming the accuracy and reliability of the developed rule to predict the outcome of future patients with Stage I endometrial adenocarcinoma.     

Puces à ADN et prédiction de l'évolution clinique des cancers ovariens

Despite debulking surgery and taxane/platinum-based chemotherapy, ovarian cancer is the most lethal pelvic gynaecological cancer in western countries, with a 25% 5-years survival. Current histo-clinical prognostic factors are insufficient to capture the heterogeneous clinical outcome of patients. A better molecular characterization of the disease is crucial to refine the prognostic classifications and to identify new therapeutic targets. DNA microarrays, which allow the quantitative measurement of expression level of the whole genome simultaneously in a single tumor sample, have been recently used towards this objective with promising results. Here, we present and discuss the main published studies and the issues to address in the future to allow the expected transfer to clinical practice.     

A prognostic test for adenocarcinoma of the lung from gene expression profiling data.

Until recently, it has been impossible to determine which patients with resected stage I lung cancer are among the 30% who will die of metastatic cancer within 5 years of surgery. Bioinformatics tools applied to lung cancer expression profiling data have identified prognostic genes that have been used to develop predictor models, but thus far, these models have not been incorporated into routine clinical use because of their inherent complexity and requirement for relatively large numbers of genes. We have used ratios of gene expression to overcome these limitations. Here, we evaluate the ability of this technique to identify patients with stage I lung adenocarcinoma at risk for recurrence. We derived candidate ratio-based tests from analysis of 36 stage I lung adenocarcinoma samples using previously published expression profiling data. Eleven of these tests were identified for additional study and assessed for classification accuracy in an independent set of 60 stage I lung adenocarcinoma samples. We then evaluated the optimal ratio-based test in the independent samples using Kaplan-Meier survival analysis. Finally, we examined the ability of this test to predict outcome in a set of 97 stage I breast adenocarcinoma. We found that subsets of the independent lung cancer samples predicted to be associated with either good or poor outcome using the optimal ratio-based test differed significantly (P=0.0056) in terms of survival with a classification accuracy of 74% (P=0.0043, Fisher's exact test). When this test was applied to stage II and III lung cancers, most specimens were classified as poor outcome cancers. Interestingly, we found that the same test significantly (P=0.0417) predicted recurrence of stage I breast tumors, suggesting that at least some of the marker genes we identified may have generalized prognostic significance for adenocarcinoma. Our results provide additional evidence that expression ratios are highly accurate in predicting cancer recurrence and may be used in a simple test to predict response to surgical therapy in early-stage lung adenocarcinoma.     

Prognostic factors in malignant mesothelioma

Many prognostic factors have been described in malignant mesothelioma, but only a limited number has been validated by prospective, confirmatory studies. These include performance status, histology, weight loss and leucocyte count. Their value as predictors of patient survival in malignant mesothelioma is limited, as is their role in current clinical practice. At this moment their main importance is their application in clinical studies, which allows a better delineation of the patient group to which the study results can be applied to and a more accurate comparison between different studies. The biological prognostic factors need confirmation by larger prospective trials and should be analysed with respect to the existing prognostic factors. In the future, their use might not be limited to the prediction of the patient survival but direct the development of new therapies and extend to the molecular prediction of tumour response.     

Angiogenesis in non-small cell lung cancer: the prognostic impact of neoangiogenesis and the cytokines VEGF and bFGF in tumours and blood

BACKGROUND: Due to a dismal prognosis of advanced lung cancer, novel screening tools and more effective treatments are clearly needed. Lately, an increasing number of tumour-released angiogenic cytokines which affect vessel formation, tumour growth, invasion, and metastasis have been identified. Vascular endothelial growth factors (VEGFs) and basic fibroblast growth factor (bFGF) are among the most important angiogenic factors. Based on available literature, we have explored the mechanisms of angiogenesis and its prognostic significance in non-small cell lung cancer, estimated by microvessel density (MVD) and the presence of VEGF and bFGF in the tumour and blood from NSCLC patients. METHODS: Several comprehensive Pubmed searches for the period January 1993 to May 2005 were performed using strategic combinations of the terms non-small cell lung cancer, angiogenesis, vascular endothelial growth factor, basic fibroblast growth factor, tumour expression, microvessel density, circulating, and serum. RESULTS: NSCLC neoangiogenesis, as measured by MVD, and tumour expression of VEGF are poor prognostic factors for survival (MVD, HR 1.8-2.0; VEGF, HR 1.5). bFGF tumour expression is also associated with poor survival and more aggressive disease. When evaluating the prognostic impact of elevated VEGF levels in blood, 10 of 16 studies (63%) indicated a negative prognostic impact. Of five studies on the prognostic value of circulating bFGF, three studies reported a negative prognostic impact, while one indicated bFGF as a good prognostic factor and one was inconclusive. CONCLUSION: Angiogenic factors are poor prognostic indicators for tumour aggressiveness and survival in NSCLC. Assessments of circulating levels of VEGF and possibly bFGF may be valuable future tools for treatment planning and monitoring of treatment effect and relapse. First, however, these blood tests need to be standardised and validated in large-scale prospective clinical trials.     

Prognostic factors in breast-cancer (review)

The challenge in breast cancer (BC) is to confide the heterogenous nature of BC and to find out the best individual therapeutic alternatives for clinical use. Recently a wide scale of sophisticated methods like morphometry, DNA flow cytometry and oncogene products have been introduced to predict the outcome of BC, but many of them are currently unsuitable for clinical use. At present limited experience in clinical practice, technical difficulties, sources of variation, and low cost-benefit ratio limit their use. Axillary lymph node status, tumour diameter and histological grade are classic and relevant predictors, although they suffer from subjectivity of assessments. Together with hormone receptor status they build the ground for clinical decision making. New demands make us to reconsider the general philosophy of the treatment of BC and there are two special groups of BCs in which we have an urgent need to find out new accurate prognostic factors. About 25% of axillary lymph node negative (ANN) tumours behave aggressively but on the other hand a group of axillary lymph node positive patients run a more favorable course than might be expected. In these situations under- or overtreatment could be avoided by accurate prediction. Recent results suggest that morphometric prognostic index, S-phase fraction and mitotic indices might be used in the prediction of BC in these situations since the prognostic results by these methods have been more accurate than by conventional prognostic methods.     

Prognostic factors in cancer

Diagnosis, prognosis, and treatment are the three core elements of the art of medicine. Modern medicine pays more attention to diagnosis and treatment but prognosis has been a part of the practice of medicine much longer than diagnosis. Cancer is a heterogeneous group of disease characterized by growth, invasion and metastasis. To plan the management of an individual cancer patient, the fundamental knowledge base includes the site of origin of the cancer, its morphologic type, and the prognostic factors specific to that particular patient and cancer. Most prognostic factors literature describes those factors that directly relate to the tumor itself. However, many other factors, not directly related to the tumor, also affect the outcome. To comprehensively represent these factors we propose three broad groupings of prognostic factors: 'tumor'-related prognostic factors, 'host'-related prognostic factors, and 'environment'-related prognostic factors. Some prognostic factors are essential to decisions about the goals and choice treatment, while others are less relevant for these purposes. To guide the use of various prognostic factors we have proposed a grouping of factors based on their relevance in everyday practice; these comprise 'essential,' 'additional,' and 'new and promising factors.' The availability of a comprehensive classification of prognostic factors assures an ordered and deliberate approach to the subject and provide safeguard against skewed approaches that may ignore large parts of the field. The current attention to tumor factors has diminished the importance of 'patient' (i.e., 'host'), and almost completely overshadows the importance of the 'environment'. This ignores the fact that the latter presents the greatest potential for immediate impact. The acceptance of a generic prognostic factor classification would facilitate communication and education about this most important subject in oncology.     

The Nottingham Prognostic Index for Invasive Carcinoma of the Breast

A useful prognostic factor in breast cancer has key roles, including identification of a group of patients whose prognosis is so good they do not require further treatment, such as adjuvant systemic therapy, after local surgery, and secondly a group with a poor prognosis for whom additional treatment would be appropriate. To be of clinical use, prognostic factors must show a wide separation in the outcome of the groups identified and select adequate numbers in each group. No single prognostic factor in invasive carcinoma of the breast satisfies all these criteria. However, the Nottingham prognostic index (NPI), which combines nodal status, tumour size and histological grade, does satisfy these criteria. The NPI has been validated by further studies in Nottingham and by studies in several other countries. Predictive factors, such as oestrogen receptor and HER-2 status, predict whether a tumour is likely to respond to a treatment, and are complimentary to prognostic factors. The NPI can be used in combination with predictive factors to select patients for systemic adjuvant treatments. There is the potential to improve the NPI by inclusion of other factors, such as vascular invasion, but any such alterations would require further validation.     

Chromosome 17 alterations identify good-risk and poor-risk tumors independently of clinical factors in medulloblastoma

Current risk stratification schemas for medulloblastoma, based on combinations of clinical variables and histotype, fail to accurately identify particularly good- and poor-risk tumors. Attempts have been made to improve discriminatory power by combining clinical variables with cytogenetic data. We report here a pooled analysis of all previous reports of chromosomal copy number related to survival data in medulloblastoma. We collated data from previous reports that explicitly quoted survival data and chromosomal copy number in medulloblastoma. We analyzed the relative prognostic significance of currently used clinical risk stratifiers and the chromosomal aberrations previously reported to correlate with survival. In the pooled dataset metastatic disease, incomplete tumor resection and severe anaplasia were associated with poor outcome, while young age at presentation was not prognostically significant. Of the chromosomal variables studied, isolated 17p loss and gain of 1q correlated with poor survival. Gain of 17q without associated loss of 17p showed a trend to improved outcome. The most commonly reported alteration, isodicentric chromosome 17, was not prognostically significant. Sequential multivariate models identified isolated 17p loss, isolated 17q gain, and 1q gain as independent prognostic factors. In a historical dataset, we have identified isolated 17p loss as a marker of poor outcome and 17q gain as a novel putative marker of good prognosis. Biological markers of poor-risk and good-risk tumors will be critical in stratifying treatment in future trials. Our findings should be prospectively validated independently in future clinical studies.     

Proposal of a scoring scale as a survival predictor in intracranial oligodendrogliomas

SummaryIntroduction Histological, clinical and radiological features, and molecular genetic analysis are among the factors that have been considered in defining the prognosis of oligodendrogliomas (OD), but they have yielded conflicting results.The purpose of this study was to test out a scoring scale based on clinical, radiological, pathological and molecular features.Material and method To identify factors with prognostic significance, we analyzed 87 treated patients with a histological diagnosis of OD. Of the parameters analyzed, age, onset, clinical status, radiological enhancement, histological necrosis, mitosis and chromosomal anomalies emerged as significant prognosis factors using univariate analysis. Multivariate analysis revealed age and chromosomal anomalies as independent factors of survival.Results The factors with a significant prognostic value were combined to determine which grouping factors best predict outcome. The proposed score is a pure number resulting from a combination of: 2 major factors: age and chromosomal anomalies (scored 3–0); 5 minor factors: onset, clinical examination, necrosis, mitoses (scored 1–0), and radiological enhancement (scored 2–0). According to our scale, 10 survival curves were produced for overall survival. Recursive partitioning of patients with the nearest score and outcome produced four groups with a significant difference in survival (p=10⁻⁵). The power of both the scale and the partitioned groups for predicting outcome was more accurate than the WHO and St Anne grading systems, and the molecular sub-classification.Conclusions Our scale is a plausible way of classifying patients harboring intracranial OD according to expected survival.     

The hypoxic tumour microenvironment, patient selection and hypoxia-modifying treatments

Tumour hypoxia has been found to be a characteristic feature in many solid tumours. It has been shown to decrease the therapeutic efficacy of radiation treatment, surgery and some forms of chemotherapy. Successful approaches have been developed to counteract this resistance mechanism, although usually at the cost of increased short- and long-term side-effects. New methods for qualitative and quantitative assessment of tumour oxygenation have made it possible to establish the prognostic significance of tumour hypoxia. The ability to determine the degree and extent of hypoxia in solid tumours is not only important prognostically, but also in the selection of patients for hypoxia-modifying treatments. To provide the best attainable quality of life for individual patients it is of increasing importance that tools be developed that allow a better selection of patients for these intensified treatment strategies. Several genes and proteins involved in the response to hypoxia have been identified as potential candidates for future use in predictive assays. Although some markers and combinations have shown potential benefit and are associated with treatment outcome, their clinical usefulness needs to be validated in prospective trials. A review of published studies was carried out, focusing on the assessment of tumour hypoxia, patient selection and the possibilities to overcome hypoxia during treatment.     

Head-to-head comparison of the 70-gene signature versus the 21-gene assay: cost-effectiveness and the effect of compliance

Both the 70-gene signature and the 21-gene assay are novel prognostic tests used to guide adjuvant chemotherapy decisions in patients with early breast cancer. Although the results of ongoing prospective trials will only become available in some years, the tests have already been included in clinical guidelines such as St. Gallen’s. In literature, the cost-effectiveness (CE) of both tests as compared to conventional prognostic tests has been described. We report on a direct comparison of CE; as different compliance rates were reported, we also taken these into account. A Markov decision model with a time horizon of 20 years was developed to assess the effects, costs and CE of three alternatives; 21-gene, 70-gene, and St. Gallen (SG) or Adjuvant Online (AO), dependent on the dataset used in patients with early, node-negative, breast cancer. Sensitivity and specificity were based on two datasets, incorporating compliances rates based on literature. For both datasets, whereas the 70-gene signature yielded more quality adjusted life years (QALYs) and was less costly; the 21-gene amounted more life years (LYs) but was more costly. The decision uncertainty surrounding the probability of CE of the Thomassen-series amounted 55% for both cost/LY and cost/QALY, for the Fan-series 80% for LY and 65% for QALYs. Taking reported compliance with discordant test results into account, in general, the effect of all strategies decreased, while the costs increased, without relatively influencing the CEA performance. This comparison indicates that the performances of the 70-gene and the 21-gene based on reported studies are close. The 21-gene has the highest probability of being cost-effective when focusing on cost/LY, while focusing on cost/QALY, the 70-gene signature was most cost-effective. The level of compliance can have serious impact on the CE. With additional data, preferably from head-to-head outcome studies and especially on compliance concerning discordant test results, calculations can be made with higher degrees of certainty.