Comparison of nomograms with other methods for predicting outcomes in prostate cancer: a critical analysis of the literature

PURPOSE: Accurate estimates of risk are essential for physicians if they are to recommend a specific management to patients with prostate cancer. Accurate risk estimates are also required for clinical trial design, to ensure homogeneous patient groups. Because there is more than one model available for prediction of most outcomes, model comparisons are necessary for selection of the best model. We describe the criteria based on which to judge predictive tools, describe the limitations of current predictive tools, and compare the different predictive methodologies that have been used in the prostate cancer literature. EXPERIMENTAL DESIGN: Using MEDLINE, a literature search was done on prostate cancer decision aids from January 1966 to July 2007. RESULTS: The decision aids consist of nomograms, risk groupings, artificial neural networks, probability tables, and classification and regression tree analyses. The following considerations need to be applied when the qualities of predictive models are assessed: predictive accuracy (internal or ideally external validation), calibration (i.e., performance according to risk level or in specific patient subgroups), generalizability (reproducibility and transportability), and level of complexity relative to established models, to assess whether the new model offers advantages relative to available alternatives. Studies comparing decision aids have shown that nomograms outperform the other methodologies. CONCLUSIONS: Nomograms provide superior individualized disease-related risk estimations that facilitate management-related decisions. Of currently available prediction tools, the nomograms have the highest accuracy and the best discriminating characteristics for predicting outcomes in prostate cancer patients.     

Nomograms for clinically localized prostate cancer. Part I: radical prostatectomy

Many nomograms are currently available for patients' and physicians' use for prediction of pathologic stage based on preoperative parameters, such as prostate-specific antigen (PSA) level, clinical stage (tumor, node, metastasis), and Gleason score from prostate biopsy specimen. Based on the probability of final pathologic stage as well as patient comorbidity and life expectancy, patients and physicians can decide whether definitive local therapy, systemic therapy, or palliative therapy would be most appropriate. Nomograms have also been developed based on preoperative parameters for prediction of biochemical recurrence-free survival outcome following surgery. These nomograms can help patients understand the long-term cancer cure rates after radical prostatectomy.     

Nomograms as predictive models

Nomograms are valuable tools for estimating the likelihood of cancer being diagnosed, the pathologic features of a localized cancer, and the prognosis of a patient after treatment. Although the available nomograms are reasonably accurate, better predictive factors including additional clinical factors and new molecular analyses are needed to improve the accuracy or predictions. Nomogram performance will also be enhanced with larger datasets of patients and longer follow-up. We review the concepts of risk stratification and the development and use of nomograms as predictive tools.     

A study of pretreatment nomograms to predict pathological stage and biochemical recurrence after radical prostatectomy for clinically resectable prostate cancer in Japanese men

BACKGROUND: Accurate pretreatment identification of the risks that prostate cancer has extended beyond the gland and that it will recur would significantly influence practice patterns. Preoperative nomograms to predict such risks have not been developed for the oriental male population. METHODS: Construction of nomograms to predict preoperatively pathological outcome and early biochemical failure following radical prostatectomy in Japanese males was based on logistic regression analysis, with predicted probabilities and 95% confidence intervals for the final model being obtained by repeating the analysis on 1000 bootstrap samples from the original cohort. RESULTS: Prostate-specific antigen level, clinical stage and biopsy Gleason score contributed significantly to the prediction of pathological stage and of biochemical failure in the univariate analysis (p < 0.001). Combined use of these three variables predicted these treatment outcomes better than any single variable (p < 0.001). Nomograms combining these three variables to predict final pathological findings and early biochemical failure were then developed. The medians and 95% confidence intervals of the predicted probabilities are presented in the nomograms. CONCLUSIONS: This information enables clinicians to use their nomograms when counseling Japanese patients, leading to more informed treatment decisions and helping to identify those with a high risk of early biochemical failure. The nomograms may also be used to assure comparability of different treatment modalities in investigational trials.     

Development and external validation of nomograms predictive of response to radiation therapy and overall survival in nasopharyngeal cancer patients

IntroductionLarge variability in the clinical outcomes has been observed among the nasopharyngeal cancer (NPC) patients with the same stage receiving similar treatment. This suggests that the current Tumour-Node-Metastasis staging systems need to be refined. The nomogram is a useful predictive tool that integrates individual variables into a statistical model to predict outcome of interest. This study was to design predictive nomograms based on the clinical and pathological features of patients with NPC.Materials and methodsClinical data of 270 NPC patients who underwent definitive radiation therapy (RT) alone or concurrent with chemotherapy were collected. Factors predictive of response to RT and overall survival (OS) were determined by univariate and multivariate analyses, and predictive nomograms were created. Nomograms were validated externally by assessing discrimination and calibration using an independent data set (N = 122).ResultsThree variables predictive of response to RT (age, histology classification and N classification) and four predictive of OS (age, performance status, smoking status and N classification), in addition to T classification, were extracted to generate the nomograms. The nomograms were validated externally, which showed perfect correlation with each other.ConclusionThe designed nomograms proved highly predictive of response to RT and OS in individual patients, and could facilitate individualised and personalised patients’ counselling and care.     

Nomograms and instruments for the initial prostate evaluation: the ability to estimate the likelihood of identifying prostate cancer

As a result of prostate cancer screening programs, approximately 10% of otherwise healthy men will be found to have an elevated prostate-specific antigen (PSA) level and therefore be at risk for harboring prostate cancer. Patients with an elevated PSA level have a wide variation in the risk for having prostate cancer diagnosed by transrectal ultrasound (TRUS)-guided prostate biopsy. To adequately counsel these patients, some form of individualized risk assessment must be given. There are several tables, artificial neural network (ANN) models, and nomograms that are available to stratify an individual patients risk for having prostate cancer diagnosed by a TRUS biopsy, either initially or on subsequent biopsies after a previous negative biopsy. Presently, nomograms are also being developed to predict the risk not only for having prostate cancer but also for clinically significant prostate cancer. The difficulty in calculating this risk for an individual patient is that the multiple competing clinical and pathologic factors have varying degrees of effect on the overall risk. This problem of competing risk factors can be overcome by the use of nomograms or ANNs. This article reviews the available instruments that are available to the urologist to enable prediction of the risk for having prostate cancer diagnosed by TRUS-guided prostate biopsy.     

Preoperative and postoperative nomograms incorporating surgeon experience for clinically localized prostate cancer

BACKGROUND:

Accurate preoperative and postoperative risk assessment has been critical for counseling patients regarding radical prostatectomy for clinically localized prostate cancer. In addition to other treatment modalities, neoadjuvant or adjuvant therapies have been considered. The growing literature suggested that the experience of the surgeon may affect the risk of prostate cancer recurrence. The purpose of this study was to develop and internally validate nomograms to predict the probability of recurrence, both preoperatively and postoperatively, with adjustment for standard parameters plus surgeon experience.

METHODS:

The study cohort included 7724 eligible prostate cancer patients treated with radical prostatectomy by 1 of 72 surgeons. For each patient, surgeon experience was coded as the total number of cases conducted by the surgeon before the patient's operation. Multivariable Cox proportional hazards regression models were developed to predict recurrence. Discrimination and calibration of the models was assessed following bootstrapping methods, and the models were presented as nomograms.

RESULTS:

In this combined series, the 10‐year probability of recurrence was 23.9%. The nomograms were quite discriminating (preoperative concordance index, 0.767; postoperative concordance index, 0.812). Calibration appeared to be very good for each. Surgeon experience seemed to have a quite modest effect, especially postoperatively.

CONCLUSIONS:

Nomograms have been developed that consider the surgeon's experience as a predictor. The tools appeared to predict reasonably well but were somewhat little improved with the addition of surgeon experience as a predictor variable. Cancer 2009. © 2009 American Cancer Society.     

Can Nomograms Predict Non–Sentinel Lymph Node Metastasis After Neoadjuvant Chemotherapy in Sentinel Lymph Node–Positive Breast Cancer Patients?

BackgroundThe predictive probability of breast cancer nomograms for non–sentinel node metastases (NSLNM) after neoadjuvant chemotherapy (NCT) in patients with a positive sentinel lymph node (SLN) biopsy is unknown. The aim of this study was to evaluate the accuracy of 3 different nomograms in patients receiving NCT.Patients and MethodsBetween 1999 and 2007, 54 patients presented with clinically N0 disease received NCT. Nomograms developed by Memorial Sloan-Kettering Cancer Center (MSKCC), Stanford University, and Tenon Hospital were used to calculate the probability of NSLNM by using tumor size at presentation and after NCT for the same patient. The discrimination of the nomograms was assessed by calculating the area under (AUC) the receiver operating characteristic curve, and it was accepted that AUC values 0.7-0.8 represent considerable discrimination.ResultsThe median patient age was 50.9 years (range, 29–67 years). Twenty-two patients (38.8%) had positive NSLNM. The MSKCC and the Stanford nomograms yielded similar AUC regardless of whether initial or post-NCT tumor size was used to determine predicted probability of NSLNM (AUCs were < 0.70). AUC was 0.74 for the Tenon model using tumor size at presentation. After NCT, the AUCs were 0.64, 0.57, and 0.78 for the MSKCC, the Stanford, and the Tenon nomograms, respectively.ConclusionAlthough the AUC of the Tenon model was acceptable for accuracy, we found a lower rate for predicting negative NSLNM in our group than in the Tenon Hospital report. All of the nomograms developed for use in the non-NCT population need to be used with caution in the NCT population     

Algorithms for prostate-specific antigen recurrence after treatment of localized prostate cancer

Nomograms are devices that predict outcome probabilities for the individual patient. Herein we discuss their strengths and limitations, focusing on responses to several frequently asked questions about nomograms. We believe these tools are useful and necessary for patient counseling, follow-up scheduling, and clinical trial design and analysis.     

Pretreatment nomogram that predicts 5-year probability of metastasis following three-dimensional conformal radiation therapy for localized prostate cancer.

PURPOSE: There are several nomograms for the patient considering radiation therapy for clinically localized prostate cancer. Because of the questionable clinical implications of prostate-specific antigen (PSA) recurrence, its use as an end point has been criticized in several of these nomograms. The goal of this study was to create and to externally validate a nomogram for predicting the probability that a patient will develop metastasis within 5 years after three-dimensional conformal radiation therapy (CRT). PATIENTS AND METHODS: We conducted a retrospective, nonrandomized analysis of 1,677 patients treated with three-dimensional CRT at Memorial Sloan-Kettering Cancer Center (MSKCC) from 1988 to 2000. Clinical parameters examined were pretreatment PSA level, clinical stage, and biopsy Gleason sum. Patients were followed until their deaths, and the time at which they developed metastasis was noted. A nomogram for predicting the 5-year probability of developing metastasis was constructed from the MSKCC cohort and validated using the Cleveland Clinic series of 1,626 patients. RESULTS: After three-dimensional CRT, 159 patients developed metastasis. At 5 years, 11% of patients experienced metastasis by cumulative incidence analysis (95% CI, 9% to 13%). A nomogram constructed from the data gathered from these men showed an excellent ability to discriminate among patients in an external validation data set, as shown by a concordance index of 0.81. CONCLUSION: A nomogram with reasonable accuracy and discrimination has been constructed and validated using an external data set to predict the probability that a patient will experience metastasis within 5 years after three-dimensional CRT.     

Algorithms for Prostate-Specific Antigen Recurrence After Treatment of Localized Prostate Cancer

Nomograms are devices that predict outcome probabilities for the individual patient. Herein we discuss their strengths and limitations, focusing on responses to several frequently asked questions about nomograms. We believe these tools are useful and necessary for patient counseling, follow-up scheduling, and clinical trial design and analysis.     

Comparisons of nomograms and urologists' predictions in prostate cancer

When applying nomograms to a clinical setting it is essential to know how their predictions compare with clinicians'. Comparisons exist outside of the prostate cancer literature. We reviewed these comparisons and conducted 2 experiments comparing predictions of clinicians with prostate cancer nomograms. By using Medline, we searched studies from January 1966 to July 1999 that compared human predictions with nomogram predictions. Next, we conducted 2 experiments: (1) 17 urologists were presented with 10 case vignettes and asked to predict the 5-year recurrence-free probabilities for each patient; (2) case presentations of 63 prostate cancer patients (including full clinical histories with complete diagnostic data and surgical findings) were made to a group of 25 clinicians who were asked to predict organ-confined disease. We found 22 published studies comparing human experts with nomograms, greater than half (13 of 22) showed the nomogram performing above the level of the human expert. Our first experiment showed urologist modification of 165 nomogram predictions led to a decrease in prediction accuracy (c-index decreased from.67 to.55, P <.05). In our second experiment, clinician predictions of organ-confined disease were comparable to the nomogram (area under the receiver operating characteristic curve [AUC] 0.78 and 0.79, respectively). A mixed-model suggests the nomogram did not augment clinician prediction accuracy (doctor excess error 1.4%, P =.75, 95% confidence interval [CI]: -10.9% to 8.2%). Our data suggest that nomograms do not seem to diminish predictive accuracy and they may be of significant benefit in certain clinical decision making settings.     

Nomograms for predicting survival and recurrence in patients with adenoid cystic carcinoma. An international collaborative study

BackgroundDue to the rarity of adenoid cystic carcinoma (ACC), information on outcome is based upon small retrospective case series. The aim of our study was to create a large multiinstitutional international dataset of patients with ACC in order to design predictive nomograms for outcome.MethodsACC patients managed at 10 international centers were identified. Patient, tumor, and treatment characteristics were recorded and an international collaborative dataset created. Multivariable competing risk models were then built to predict the 10 year recurrence free probability (RFP), distant recurrence free probability (DRFP), overall survival (OS) and cancer specific mortality (CSM). All predictors of interest were added in the starting full models before selection, including age, gender, tumor site, clinical T stage, perineural invasion, margin status, pathologic N-status, and M-status. Stepdown method was used in model selection to choose predictive variables. An external dataset of 99 patients from 2 other institutions was used to validate the nomograms.FindingsOf 438 ACC patients, 27.2% (119/438) died from ACC and 38.8% (170/438) died of other causes. Median follow-up was 56 months (range 1–306). The nomogram for OS had 7 variables (age, gender, clinical T stage, tumor site, margin status, pathologic N-status and M-status) with a concordance index (CI) of 0.71. The nomogram for CSM had the same variables, except margin status, with a concordance index (CI) of 0.70. The nomogram for RFP had 7 variables (age, gender, clinical T stage, tumor site, margin status, pathologic N status and perineural invasion) (CI 0.66). The nomogram for DRFP had 6 variables (gender, clinical T stage, tumor site, pathologic N-status, perineural invasion and margin status) (CI 0.64). Concordance index for the external validation set were 0.76, 0.72, 0.67 and 0.70 respectively.InterpretationUsing an international collaborative database we have created the first nomograms which estimate outcome in individual patients with ACC. These predictive nomograms will facilitate patient counseling in terms of prognosis and subsequent clinical follow-up. They will also identify high risk patients who may benefit from clinical trials on new targeted therapies for patients with ACC.FundingNone.     

Multicenter validation of two nomograms to predict non-sentinel node involvement in breast cancer

Introduction

Nomograms are used to predict the involvement of non-sentinel nodes (nSN) in breast cancer. This study attempts to externally validate two of the more commonly used nomograms (MSKCC and Stanford University).

Materials and methods

Five hundred and one cases of positive SNB with posterior axillary lymphadenectomy from 11 Spanish hospitals with widespread experience of the technique were studied. In all cases, an estimate of the probability of nSN involvement was made using the MSKCC and the Stanford University nomograms. Discrimination was assessed by calculating the area under the receiver operating characteristic curve. To assess the calibration of the nomogram, observed probability was plotted against the nomogram-calculated predicted probability.

Results

The overall predictive accuracy of the MSKCC nomogram was 0.684 (95 % confidence interval, 0.635–0.732), while in the case of that from Stanford the predictive accuracy was 0.658 (95 % confidence interval 0.607–0.709). The mean predicted probability of nSN metastases in each group of patients was correlated with the observed probability with an acceptable concordance (r = 0.820; p < 0.004 in MSKCC nomogram and r = 0.888; p < 0.001 in Stanford nomogram).

Conclusion

These nomograms can be useful tools in the evaluation of patients with breast cancer and positive sentinel nodes but other factors, including a comprehensive clinical assessment, must be used to decide the most appropriate surgical approach for an individual patient, especially with regard to avoiding unnecessary lymphadenectomy.     

Evaluation of quality of life for diverse patient populations

Summary Quality of life is used increasingly as a primary and secondary endpoint of clinical investigations of new therapies. Quality of life information may be especially useful for the assessment of cancer treatments, where increases in survival may be accompanied by detrimental side effects. The recognition of the importance of quality of life has led to the recent proliferation of cancer specific quality of life instruments. As more is understood about the heterogeneity of patient populations, however, we must assess how culturally defined factors may impact patient quality of life and its assessment. Quality of life instruments are diverse, ranging from those focusing on objective measures of functionality to those assessing subjective measures of patient preferences for their current health state. These instruments have been developed for use in the general population and for disease-specific populations. Assessment of the appropriateness of potential quality of life instruments in specific clinical settings, in addition to understanding the cultural diversity of the clinical population being studied, will guide the researcher in the choice of an appropriate quality of life instrument for cancer clinical trials.     

Prediction tools in surgical oncology

Artificial neural networks, prediction tables, and clinical nomograms allow physicians to transmit an immense amount of prognostic information in a format that exhibits comprehensibility and brevity. Current models demonstrate the feasibility to accurately predict many oncologic outcomes, including pathologic stage, recurrence-free survival, and response to adjuvant therapy. Although emphasis should be placed on the independent validation of existing prediction tools, there is a paucity of models in the literature that focus on quality of life outcomes. The unification of tools that predict oncologic and quality of life outcomes into a comparative effectiveness table will furnish patients with cancer with the information they need to make a highly informed and individualized treatment decision.     

Problems of the preoperative prediction of the pathological stage in prostate cancer

The pathological stage of the tumor is the most influential prognostic factor for progression after radical prostatectomy. However, as many as 50% of men undergoing radical prostatectomy are found to have extraprostatic disease in the pathological specimen. Accurate identification of the risks of disease extension and of disease recurrence prior to radical prostatectomy would thus be useful in counseling men presenting with clinically localized prostate cancer. Nomograms may help patients and physicians make more informed treatment decisions based on the probability of pathological stage. Partin and co-workers popularized the use of a pretreatment nomogram based on PSA (prostate specific antigen), clinical stage (TNM stage) and biopsy Gleason score to predict the pathological stage of localized prostate cancer. However, it may not be directly applicable to Japanese males, and the interpretation and comparison of data sets should be done with caution and careful consideration. Although attempts have been made to establish a nomogram for Japanese patients, been tried, it is still based on the data for a small number of patients. More data from a greater number of patients and validation analysis are essential. Recently, artificial neural networks (ANN) have been shown to be effective in predicting pathologic stage in men with clinically localized prostate cancer. The use of ANNs is a relatively new concept and the data is based on Western people; thus, the data analysis for Japanese patients is necessary. The present paper mainly outlines the usefulness and problems for the preoperative prediction of the pathological stage in prostate cancer by nomograms and artificial neural networks.     

Nomograms for clinically localized prostate cancer. Part II: radiation therapy

Prostate cancer can be effectively treated with either external beam radiation techniques or with brachytherapy. This study was designed to address the methodology that is used to assess outcome data in the current radiation literature and to evaluate available nomograms that can be used to predict outcomes. A literature search was performed and 12 articles reviewed. Risk stratification was the most frequently used methodology to analyze data. This method encompasses disease-specific variables: the pretreatment prostate-specific antigen (PSA) value and the Gleason score are classified by using cut points into low, intermediate, and high-risk groups. Another methodology uses nomograms to predict outcome. The nomogram uses continuous values of each variable so that the outcome probability for a specific set of parameters is quite specific. The advantage of nomogram analysis over risk stratification analysis is presented. In conclusion, only 3 reports were identified in the radiation literature that used a nomogram to predict outcome. One of the nomograms is proprietary and difficult to interpret. The other 2 nomograms, 1 for 3-dimensional radiation and the other for brachytherapy, have been incorporated into hand-held devices that can be used at consultation with the patient to discuss outcome probabilities to assist in treatment decisions.     

An updated catalog of prostate cancer predictive tools

Accurate estimates of risk are essential for physicians if they are to recommend a specific management to patients with prostate cancer. Accurate risk estimates also are required for clinical trial design to ensure that homogeneous, high-risk patient groups are used to investigate new cancer therapeutics. Using the MEDLINE database, a literature search was performed on prostate cancer predictive tools from January 1966 to July 2007. The authors recorded input variables, the prediction form, the number of patients used to develop prediction tools, the outcome being predicted, prediction tool-specific features, predictive accuracy, and whether validation was performed. Each prediction tool was classified into patient clinical disease state and the outcome being predicted. First, the authors described the criteria for evaluation (predictive accuracy, calibration, generalizability, head-to-head comparison, and level of complexity) and the limitations of current predictive tools. The literature search generated 109 published prediction tools, including only 68 that had undergone validation. An increasing number of predictive tools addressed important endpoints, such as disease recurrence, metastasis, and survival. Despite their limitations and the limitations of data, predictive tools are essential for individualized, evidence-based medical decision making. Moreover, the authors recommend wider adoption of risk-prediction models in the design and implementation of clinical trials. Among prediction tools, nomograms provide superior, individualized, disease-related risk estimations that facilitate management-related decisions. Nevertheless, many more predictive tools, comparisons between them, and improvements to existing tools are needed.