Home 
Prostate Cancer European Urology - Nomograms and Medicine
Prostate Cancer European Urology - Nomograms and Medicine
 |
| European Urology - Nomograms and Medicine |
|
|
|
|
|
| Wednesday, 25 October 2006 | ||
|
Volume 50, Issue 5, Pages 884-886 (November 2006) Article Outline: Prostate cancer is the most common malignancy in European men. In 2002, an estimated 225,000 men were newly diagnosed with prostate cancer and about 83,000 died from this disease in Europe (http://www-dep.iarc.fr) [1]. In the United States, this cancer is the most common solid malignancy in men, with an annual incidence rate of 180cases/100,000, and growing at a 2.3% rate since 1992 [2]. The increase in prostate cancer incidence in the United States is reflected by the approval from the Food and Drug Administration for the use of serum prostate-specific antigen (PSA) as a screening test. Although the survival benefits of PSA screening remain the subject of prospective studies in Western nations [3], the consequent rate of metastatic prostate cancer began to decrease at a dramatic annual rate of 18% in 1991 and now represents <5% of newly diagnosed tumours. In 2006, a decrease in the cancer mortality rankings was observed in the United States. This phenomenon has been driven by a 4% yearly drop in prostate cancer mortality rates for 10 yr [2]. Today, prostate cancers are identified at an earlier stage, which allows effective treatment with surgical or nonsurgical approaches. These paradigmatic improvements in cancer prognosis are easily recognised by physicians and conveyed to patients by powerful instruments called nomograms. Strictly speaking, within the medical field, a nomogram is a graphic calculating device, a two-dimensional diagram designed to allow the approximate graphic computation of the likelihood of a clinical event. The origin of nomograms goes back to the mid-1770s in the statistical atlases of William Playfair, the indicator diagrams of James Watt, and the writings of Johann Heinrich Lambert [4]. However, the first medical nomogram, presented in 1928, is credited to L.J. Henderson, a former physiologist and president of the History of Science Society. Henderson found that the only way he could describe a chemical system as complicated as blood was by a diagram called a “nomogram” [5]. Today, there are >1700 publications related to nomograms registered in Medline, with 1100 of these coming from the 1990s with the advent and evolution of the digital age, particularly the S programing language that is the basis for statistical, graphic applications (S+ or R) that holds libraries of functions such as Design—enabling graphic translations of mathematical complex equations. For patients with prostate cancer, the paradigm on prognosis after treatment was changed a decade ago with seminal reports by Kattan et al. [6], [7] because the treatment decision-making process very commonly incorporates these formidable practical tools. Furthermore, from the date of their original publication to the date of this writing, the medical literature has benefitted from >760 studies that have expanded the connection between nomogram applications and disease entities. In clinical practice, the nomogram represents the graphic translation of elaborate regression analyses. The common end results offer a statement on prognosis, with increased patient awareness of the likely course of the disease after receiving a given treatment adjusting for the characteristics that the patient currently presents. In this issue of European Urology, the review by Chun et al. [8] focuses on prostate cancer predictive tools and provides a state-of-the art review on those tools presented as nomograms to date. The innovative features of this report are represented by a succinct description of how to interpret a nomogram, how their accuracy is measured, which ones had independent validation, and cleverly grouping these tools by the predictive clinical event [8]. The main take home message for the prostate cancer spectrum relates to the fact that clinical probability is no longer assessed implicitly on the basis of the clinician's global judgement but explicitly on the basis of a prediction equation deriving from the individual clinical characteristics. Thus, nomograms created a shift in our predictive paradigms from an exclusionary, imprecise, heterogeneous yet practical and commonly used strata of grouping systems to one that provides a continuum scale of prediction. Nomographs of comprehensive prediction models focus our attention on the prediction given, enhancing the comprehension and usability of the model and not its individual components. The model can be conceived as a test. As indicated by the authors, the value measure of this model is the predictive accuracy, that is, the concordance index, the area under the curve, a far cry from a single variable cut, hazard, or odds ratios. This prediction ranges from 0.5 (coin flip) to 1 (perfect prediction). Most of the predictive accuracy falls within the range, the farther from 1 the more limited; yet, we can easily quantify this limitation. Ideally, these models must be validated by measuring their reliability and accuracy in an independent population. If the model is validated, then the predictions should be universal assuming management equivalency. Yes, unfortunately, clinical events in some will escape these estimations; however, these predictions will accurately reflect the clinical course of the majority. By using the well-validated preoperative prostate nomogram [6] as an example, we can expect that the majority of patients with a serum PSA level of 6ng/ml, a Gleason score of 8, and prostate cancer staged clinically at T1c, who are treated with radical prostatectomy, will have an 83% probability of being disease free after 5 yr. Assuming surgical equivalencies, this outcome would be the likely event for most, not all, because the predictive accuracy of the model is limited at 0.76. Importantly, these predictions are not affected by regions, states, or prevalence of disease. They depend on the clinical characteristics of the individual patient. Today, nomogram popularity relates to statistical models built with simple yet powerful and practical information. For the future, superior clinical applicability is expected for those who exhibit concordance indices >0.75 after external validation. We can look forward to nomogram deployment as the selection criteria used in prospective randomised trials as they homogenise risk (inclusion or exclusion) over the traditional, arbitrary, yet “clinically” sensible practice of heterogeneous grouping. How we can improve predictive accuracy? Nomograms bring into visual perspective the effect exerted by continuous variables against measured end points. From a predictive outlook, this is tremendous because categorisation of continuous variables often comes at the expense of predictive accuracy. However, predictive accuracy can be improved by arithmetic transformation, as done with PSA, signalling the correlation between biomarker and the course of the disease, as presented in many of the models shown in the tables from the review by Chun et al. [8]. Importantly, predictive accuracy is proportional to objective data and independent of the number of variables. Nonetheless, in the digital age, we have come to the point where the tool base (probability tables [9], classification and regression tree [CART] analysis [10], and nomograms [6], [7]) and the formulas are not the main limitations in terms of predictions. The formulas can be loaded into patient care software solutions, PDAs, or Internet Web sites that will give us those predictions; nomograms can be used to explain how those models work. The two important restrictions are science and data quality. For the science, as much as we come to learn of prostate cancer biology, a wealth of information still escapes us; however, steady progress is being made and powerful biomarkers such as PSA will come along. We can, however, make a significant measurable impact on the data quality. Indeed, the entire multivariable analysis basis of the nomograms presented in the review shared a common theme—they did not come from randomised, prospective, controlled data. Retrospective assessments, thorough as they can be, carry the intrinsic selection bias that leads to the “x” or “y” procedure or intervention, selection bias that is associated with the outcome of interest. This is a major burden to defeat in an era of evidence-based medicine. In summary, nomograms have empowered patients and physicians in their fight against cancer. In the clinical setting they have provided superior individualised disease-related risk estimations. Both patients and physicians benefit from a more realistic boundary of outcome expectations to help in the management-related decisions. References 1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin. 2005;55:74–108. 2. Edwards BK, Brown ML, Wingo PA, et al.. Annual report to the nation on the status of cancer, 1975–2002, featuring population-based trends in cancer treatment. J Natl Cancer Inst. 2005;97:1407–1427. 3. Miller AB, Madalinska JB, Church T, et al.. Health-related quality of life and cost-effectiveness studies in the European randomised study of screening for prostate cancer and the US Prostate, Lung, Colon and Ovary trial. Eur J Cancer. 2001;37:2154–2160. 4. Lambert JH. Pyrometrie; oder, Vom Maasse des Feuers und der Wärme mit acht Kupfertafeln. Berlin: Bey Haude & Spener; 1779;pp. 424–88. 5. Henderson LJ. Blood: a study in general physiology. Vol. 3. New Haven, CT: Yale University Press; 1928;p. 148, Fig. 141. 6. Kattan MW, Eastham JA, Stapleton AM, Wheeler TM, Scardino PT. A preoperative nomogram for disease recurrence following radical prostatectomy for prostate cancer. J Natl Cancer Inst. 1998;90:766–771. 7. [7]Kattan MW, Wheeler TM, Scardino PT. Postoperative nomogram for disease recurrence after radical prostatectomy for prostate cancer. J Clin Oncol. 1999;17:1499–1507. 8. Chun FK-H, Karakiewicz PI, Briganti A, et al.. Prostate cancer nomograms: an update. Eur Urol. 2006;50:914–926. 9. Partin AW, Kattan MW, Subong EN, et al.. Combination of prostate-specific antigen, clinical stage, and Gleason score to predict pathological stage of localized prostate cancer. A multi-institutional update. JAMA. 1997;277:1445–1451. 10. Graefen M, Haese A, Pichlmeier U, et al.. A validated strategy for side specific prediction of organ confined prostate cancer: a tool to select for nerve sparing radical prostatectomy. J Urol. 2001;165:857–863. Fernando J. Bianco Jr. Chief Division of Urology Oncology, Department of Urology, George Washington University, 2150 Pennsylvania Avenue, NW, Suite 3-417, Washington, DC 20037, United States published online 14 August 2006.
Please log-in or register in order to submit comments. Powered by AkoComment! |
||
| < Prev | Next > |
|---|
|
UroToday, 1802 Fifth Street, Berkeley CA 94710 510.540.0930 (fax), info@urotoday.com ISSN 1939-4810
Privacy Policy | © 2009 UroToday ® All Rights Reserved |
