Printer Friendly

Accuracy capabilities comparisons between Karakiewicz, Kattan and Cindolo nomograms in predicting outcomes for renal cancer carcinoma: a systematic review and meta-analysis.

Introduction

Over the past years, the management options for patients with renal cell carcinoma (RCC) at all stages have increased. (1) Partial or total nephrectomy is the standard treatment for locally resectable tumours with curative intention. (2) However, 20% to 40% of surgically treated tumours will develop recurrence during follow-up, which underlines the importance of tailored follow-up regimens and the evaluation of effectiveness of adjuvant therapies. (3)

In this context, the use of several prognostic factors and models has gained popularity to predict outcomes of patients affected by RCC. In general, all these prognostic tools are more accurate than the standard TNM classification or Fuhrman grade in predicting survival outcomes. (4) A substantial advantage of prognostic tools is the ability to measure the predictive accuracy, which allows an objective evaluation of the performance itself. (5) Several predictive models have been proposed; however, some doubts still persist about their discriminative capabilities in predicting oncological outcomes for RCC.

To this regard, preoperative Karakiewicz, postoperative Karakiewicz, Kattan and Cindolo models have been internally and externally validated in different populations. (6-9) Limitations of nomograms include the racial difference among populations, the variability in accuracy, and their characteristics to outperform risk groups.

We review the discriminative capabilities of these four predictive models (preoperative Karakiewicz, postoperative Karakiewicz, Kattan and Cindolo models) and perform a meta-analysis to yield pooled area under the receiver operator curves (AUCs) for model comparison.

Methods

This analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analysis guidelines. (10) An electronic search of Medline and Embase was undertaken until July 2014. The search was limited to English articles. The search terms included RCC and related terms, nomogram, integrated staging systems, cancer-specific survival, disease recurrence, predictors, and outcomes.

Citation lists of retrieved articles were screened manually to ensure sensitivity of the search strategy. References of the included papers were also manually searched to identify other potential relevant studies. This meta-analysis did not include patient-level studies, but only included studies with statistically combined accuracies reporting the use of nomograms. Studies were reviewed by two independent reviewers (GIR, AD). Differences in opinion were discussed in consultation with the last author (GM).

The end points of interest were DFS and CSS. The AUC value, total number of patients, and the number of cancer-related deaths were extracted from the included references. A meta-analysis of the ROC curves was performed based on methods reported by Walter and colleagues. (11) Basically, the AUCs were converted to odds ratios (ORs) using the following equation (equation 1):

{1) AUC = OR/[(OR - 1).sup.2] [(OR - 1} - ln(OR)]

The standard error of the AUC and OR was calculated as follows:

(2) SE (AUC) = [square root of (Q1/n + Q2/m - [AUC.sup.2] (m + n)/mn)]

In this equation, Q1 = AUC/(2-AUC), Q2 = 2[AUC.sup.2]/(1 + AUC), and

(3) SE(OR) = SE(AUC)[(OR - 1).sup.3]/(OR + 1)ln(OR) - 2(OR - 1)

For the meta-analysis, ln(OR) was used for data pooling. SE[ln(OR)] was calculated through a first-order Taylor series conversion, where SE[ln(OR)] = (1/OR) x SE[OR]. Begg's and Egger's methods were used to assess publication bias. (12,13) Begg's test was based on the rank correlation between the observed effect sizes and observed standard errors, while Egger's regression intercept is similar to Begg's but used actual values instead of ranks.

Statistical heterogeneity was assessed using the CochranQ and [I.sup.2] statistics. Specifically, statistical heterogeneity was tested using the chi-square test. A value of p < 0.10 was used to indicate heterogeneity. In the case of a lack of heterogeneity, fixed-effects model was used to assess the overall combined OR. For each nomogram, the combined OR was transformed back to a converted AUC (cAUC) using equation 1. All of the tests were two-tailed, and a p < 0.05 was regarded as significant. The analyses were performed using RevMan software v.5.1 (Cochrane Collaboration, Oxford, UK).

Results

After excluding redundant literature, a total of 16 studies were identified, which included 26 710 patients (Table 1, Fig. 1). (1,4,6,7,9,14-24) In total, the preoperative Karakiewicz nomogram, postoperative Karakiewicz nomogram, Kattan nomogram, and the Cindolo nomogram were validated in 12 065, 12 868, 6036 and 4045 patients, respectively. In all of the included models, we did not observe any publication bias as assessed by the Begg's and Egger's methods (Fig. 2). The weighted median follow-up for all patients was 60 months (range: 33.6-82.0). In studies on DFS, the weighted median follow-up was 60 months (range: 37.0-81.0), while the weighted median follow-up for CSS was 55.2 months (range: 33.6-82.0). The pooled DFS for the preoperative Karakiewicz nomogram, postoperative Karakiewicz nomogram, the Kattan nomogram, and the Cindolo nomogram were 84.98%, 88.27%, and 87.07%, respectively.

The pooled CSS for the preoperative Karakiewicz nomogram, the postoperative Karakiewicz nomogram, the Kattan nomogram, and the Cindolo nomogram were 82.68%, 86.03%, 86.33%, and 84.20%, respectively.

Disease-recurrence survival

The postoperative Karakiewicz model was validated in 3 studies. Non-significant heterogeneity was found in this nomogram ([x.sup.2] = 0.19, [I.sup.2] = 0%, p = 0.91). The weighted median follow-up for all patients was 53.5 months (range: 37.0-65.0). The pooled ORs (95% confidence interval [CI]) and the corresponding cAUC value were 4.32 (1.13-16.47) and 0.728, respectively.

The Kattan model was validated in 8 studies. Non-significant heterogeneity was found in this nomogram ([x.sup.2] = 4.02, [I.sup.2] = 0%, p = 0.86). The weighted median follow-up for all patients was 60 months (range: 33.6-82.0). The pooled ORs (95% CI) and the corresponding cAUC value were 2.97 (1.66-5.34) and 0.675, respectively.

The Cindolo model was in validated in 4 studies. Non-significant heterogeneity was found in this nomogram ([x.sup.2] = 3.53, [I.sup.2] = 0%, p = 0.94). The weighted median follow-up for all patients was 60 months (range: 42.0-67.0). The pooled ORs (95% CI) and the corresponding cAUC value were 3.89 (2.06-7.34) and 0.713, respectively. The test of overall effect was statistical significant (Z = 5.92, p < 0.00001) (Fig. 3). The Mantel-Haenszel derived comparison of cAUC values revealed better predictive capability for the postoperative Karakiewicz nomogram versus the Kattan nomogram (p < 0.01), but not versus the Cindolo model (p = 0.432) and between the Cindolo versus Kattan models (p = 0.03) (Table 2).

Cancer-specific survival

The preoperative Karakiewicz model was validated in 4 studies. Non-significant heterogeneity was found in this nomogram ([chi square] = 0.40, [I.sup.2] = 0%, p = 0.94). The weighted median follow-up was 48.50 months (range: 48.0-50.4). The pooled ORs (95% CI) and the corresponding cAUC value were 8.47 (range: 2.79-25.70) and 0.81, respectively.

The postoperative Karakiewicz model was validated in 7 studies. Non-significant heterogeneity was found in this nomogram ([chi square] = 0.46, [I.sup.2] = 0%, p = 1.00).The weighted median follow-up was 57.0 months (range: 36.6-82.0). The pooled ORs (95% CI) and the corresponding cAUC value were 8.82 (range: 2.08-37.40) and 0.814, respectively.

The Kattan model was validated in 4 studies. Non-significant heterogeneity was found in this nomogram ([chi square] = 0.02, [I.sup.2] = 0%, p = 1.00).The weighted median follow-up was 62.5 months (range: 37.2-65.0). The pooled ORs (95% CI) and the corresponding cAUC value were 6.52 (range: 1.80-23.57) and 0.780, respectively.

The Cindolo model was in validated in 2 studies. Non-significant heterogeneity was found in this nomogram ([chi square] = 0.30, [I.sup.2] = 0%, p = 0.59). The weighted median follow-up was 62.5 months (range: 60.0-65.0). The pooled ORs (95% CI) and the corresponding cAUC value were 2.61 (1.58-4.30) and 0.655, respectively.

The overall weighted follow-up was 55.2 (range: 33.682.0). The test of overall effect was statistical significant (Z = 6.26, p < 0.00001) (Fig. 2). The Mantel-Haenszel derived comparison of cAUC values revealed better predictive capability for the preoperative Karakiewicz nomogram versus the Kattan nomogram (p < 0.01) and versus the Cindolo model (p < 0.01), but also between the postoperative Karakiewicz model versus the Kattan model (p < 0.01) and the Cindolo model (p < 0.01). The Kattan model showed better discriminative capability versus the Cindolo model (p < 0.01). No statistical difference was observed between both Karakiewicz models (p = 0.730) (Table 3).

Discussion

Renal cancer nomograms have been established to counsel patients before treatment. In this context, the Karakiewicz, Kattan and Cindolo models have been widely validated in different populations from different countries. (25) However, the best-performing model remains unknown.

Kattan and colleagues from the Memorial Sloan-Kettering Cancer Center developed a nomogram to predict the 5-year progression-free survival of patients undergoing radical nephrectomy for non-metastatic RCC of various histological subtypes. The four factors included in this nomogram were the presence of symptoms, histological subtype, tumour size, and standard TNM stage according to the 1997 version. (9) When applied to external populations in Europe, the original Kattan nomogram has shown variable prognostic accuracy ranging from 61% to 81%. (4,18-21,23)

In 2007, Karakiewicz and colleagues attempted to improve on the accuracy of the aforementioned models by including more variables that have traditionally been shown to predict survival among patients with RCC. The cohort on which the model was developed included over 2500 patients with various stages of RCC treated at 5 different centres. Their final model ultimately incorporated TNM stage, tumour size, histological subtype, age, sex, and symptoms at presentation to predict 1-, 2-, 5- and 10-year cancer-specific mortality. The internally validated accuracy of the nomogram was 86%, (6) but the externally accuracy reached 90.5%. (1,6,15,24)

Karakiewicz and colleagues examined the ability of T and M stages to predict freedom from cancer-specific mortality (CSM) (n = 2474). (7) In addition to T and M stages, other variables, such as age, gender, tumour size, and symptoms, resulted in an integrated staging system that provided predictions of CSM-free survival at 1, 2, 5, and 10 years after nephrectomy. Discrimination of that model ranged from 84% to 88% within an external validation cohorts. (1,7,16,17)

A second preoperative model focusing on RCC recurrence after nephrectomy was developed by Cindolo and colleagues (n = 660). (8) This staging system relied on symptoms at presentation and on preoperative tumour size. The Cindolo and colleagues nomogram's discriminatory ability ranged from 67% in European patients to 75% in Chinese patients. (4,8,22)

The diffusion of several nomograms to discriminate between similar end points is problematic. It seems obvious that the choice of one or several of these models should be based on their predictive ability and accuracy. (25)

One should also take into account that not all of these end points can be defined with certainty. For example, the recurrence-free rate could be limited by the heterogeneity of follow-up or the characteristics of the imaging techniques used. Moreover, it seems obvious that predicting mortality improves the gain in accuracy of the model itself. Based on our results, the converted cAUC values of the pooled ORs for predicting CSS were higher than those for predicting DFS. Therefore, common limitations of the models, such as racial difference among population and sample size, should be considered.

For these reasons, we performed a systematic review and meta-analysis to obtain the derived AUC from pooled ORs for each model and to compare models. We transformed the converted AUC values into ORs using methods reported by Walter and colleagues. (11)

To the best of our knowledge, this is the first meta-analysis investigating the discriminative capabilities of current nomogram for RCC and including 26 710 patients.

Our results confirmed that the preoperative Karakiewicz, postoperative Karakiewicz, and the Kattan models had a combined AUC value more than 0.70 for predicting CSS, while only the postoperative Karakiewicz model to predict DFS suggested stable discriminative capabilities in different populations.

In particular, the postoperative Karakiewicz (p < 0.01) and Cindolo (p = 0.32) models better exhibited cAUC values than the Kattan nomogram for DFS (Table 2). Regarding the discriminative capability for CSS, both Karakiewicz models showed the best predictive ability over the Kattan (all p < 0.01) and the Cindolo (all p < 0.01) models (Table 3).

Based on accuracy and pooled ORs derived from the current meta-analysis, the preoperative and postoperative Karakiewicz models have given the better predictive capability for predicting CSS (both cAUC = 0.81), while the postoperative Karakiewicz (cAUC = 0.728) was better than Cindolo and Kattan for predicting DFS (cAUC = 0.728). On the contrary, the Kattan and Cindolo models showed intermediate predictive capability in predicting CSS and DFS, respectively.

The differences in pooled OR observed between nomograms could be explained by the heterogeneity of variables included in the models itself. In fact, this may be considered a use for these nomograms. We attempted to counteract these limitations by calculating the pooled AUC of all published data.

Our study has its limitations. Firstly, the median follow-up was different among studies. Secondly, we used a new method proposed by Walter and colleagues to convert the reported AUCs to ORs for the meta-analysis. However, the precision of this conversion will be affected by the reported AUC values with varying decimal places (we used three decimal places). Moreover, the conversion formula (equation 1) (from OR to AUC) cannot be inverted analytically (from AUC to OR). Therefore, we obtained the OR through by graphing using Derive v.6 (Texas Instruments, Inc.). Furthermore, the formula is a monotonically increasing function, guaranteeing the feasibility of getting OR through this method. Moreover we did not conduct this meta-analysis at a patient level, but only statistically combined accuracies of studies using previous nomograms. It may be expected that the same patients were included in more models. However, it is impossible to discriminate this at this manuscript level.

Thirdly, although there is a low risk of publication bias, the choice of nomograms was made based on previous publications and available local data. Finally we did not evaluate possible confounding factors that could have influenced that AUC. However, this was out of the scope of the study.

We would also underline that, although these nomograms have been originally created for specific outcomes, they have also been applied for different end points. We included forest plots to evaluate the same outcome and this can be translated in the clinical practice.

Conclusion

The predictive abilities of the pre- and post-operative Karakiewicz models are higher than Kattan or Cindolo in predicting DFS and CSS. The Cindolo and the Kattan nomogram showed relatively intermediate capability for DFS and CSS, respectively, if compared to other models. These results allow us to evaluate the risk of RCC-specific recurrence and mortality before suggesting nephrectomy, partial nephrectomy or adjuvant chemotherapy after surgery.

http://dx.doi.org/10.5489/cuaj.2479

Published online June 18, 2015.

Competing interests: The authors declare no competing financial or personal interests.

This paper has been peer-reviewed.

References

(1.) Cindolo L, Chiodini P, Brookman-May S, et al. Assessing the accuracy and generalizability of the preoperative and postoperative Karakiewicz nomograms for renal cell carcinoma: Results from a multicentre European and US study. BJU Int 2013;1 12:578-84. http://dx.doi.org/10.1111/j.1464- 410X.2012.11670.x

(2.) Ljungberg B, Bensalah K, Canfield S, et al. EAU Guidelines on Renal Cell Carcinoma: 2014 Update. Eur Urol 2015;67:913-24. http://dx.doi.org/10.1016/j.eururo.2015.01.005

(3.) Hollingsworth JM, Miller DC, Daignault S, et al. Five-year suivival after surgical treatment for kidney cancer: A population-based competing risk analysis. Cancer 2007;109:1763-8. http://dx.doi.org/10.1002/cncr.22600

(4.) Cindolo L, Patard JJ, Chiodini P, et al. Comparison of predictive accuracy of four prognostic models for non-metastatic renal cell carcinoma after nephrectomy: A multicenter European study. Cancer 2005;104:136271. http://dx.doi.org/10.1002/cncr.21331

(5.) Motzer RJ, Mazumdar M, Bacik J, et al. Effect of cytokine therapy on survival for patients with advanced renal cell carcinoma. J Clin Oncol 2000;18:1928-35.

(6.) Karakiewicz PI, Briganti A, Chun FK, et al. Multi-institutional validation of a new renal cancer-specific survival nomogram. J Clin Oncol 2007;25:1316-22. http://dx.doi.org/10.1200/JC0.2006.06.1218

(7.) Karakiewicz PI, Suardi N, Capitanio U, et al. A preoperative prognostic model for patients treated with nephrectomy for renal cell carcinoma. Eur Urol 2009;55:287-95. http://dx.doi.org/10.1016/j. eururo.2008.07.037

(8.) Cindolo L, de la Taille A, Messina G, et al. A preoperative clinical prognostic model for non-metastatic renal cell carcinoma. BJU Int 2003;92:901-5. http://dx.doi.org/10.1111/j.1464- 410X.2003.04505.x

(9.) Kattan MW, Reuter V, Motzer RJ, et al. A postoperative prognostic nomogram for renal cell carcinoma. J Urol 2001;166:63-7. http://dx.doi.org/10.1016/S0022-5347(05)66077-6

(10.) Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: Explanation and elaboration. PLoS Med 2009;6:e1000100. http://dx.doi.org/10.1371/journal.pmed.1000100

(11.) Walter SD, Sinuff T. Studies reporting ROC curves of diagnostic and prediction data can be incorporated into meta-analyses using corresponding odds ratios. J Clin Epidemiol 2007;60:530-4. http://dx.doi.org/10.1016/j.jclinepi.2006.09.002

(12.) Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994;50:1088-101. http://dx.doi.org/10.2307/2533446

(13.) Egger M, Davey Smith G, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629-34. http://dx.doi.org/10.1136/bmj.315.7109.629

(14.) Liu Z, Lv J, Ding K, et al. Validation of the current prognostic models for nonmetastatic renal cell carcinoma after nephrectomy in Chinese population: A 15-year single center experience. Int J Urol 2009;16:268-73.

(15.) Karakiewicz PI, Suardi N, Capitanio U, et al. Conditional survival predictions after nephrectomy for renal cell carcinoma. J Urol 2009;182:2607-12. http://dx.doi.org/10.1016/j.juro.2009.08.084

(16.) Kutikov A, Egleston BL, Wong YN, et al. Evaluating overall survival and competing risks of death in patients with localized renal cell carcinoma using a comprehensive nomogram. J Clin Oncol 2010;28:311-7. http://dx.doi.org/10.1200/JC0.2009.22.4816

(17.) Gontero P, Sun M, Antonelli A, et al. External validation of the preoperative Karakiewicz nomogram in a large multicentre series of patients with renal cell carcinoma. World J Urol 2013;31:1285-90. http://dx.doi.org/10.1007/s00345-012-0896-z

(18.) Hupertan V, Roupret M, Poisson JF, et al. Low predictive accuracy of the Kattan postoperative nomogram for renal cell carcinoma recurrence in a population of French patients. Cancer 2006;107:2604-8. http://dx.doi.org/10.1002/cncr.22313

(19.) Suzuki K, Nishiyama T, Hara N, et al. Kattan postoperative nomogram for renal cell carcinoma: Predictive accuracy in a Japanese population. Int J Urol 2011;18:194-9.

(20.) Tan MH, Li H, Choong CV, et al. The Karakiewicz nomogram is the most useful clinical predictor for survival outcomes in patients with localized renal cell carcinoma. Cancer 2011;1 17:5314-24. http://dx.doi.org/10.1002/cncr.26193

(21.) Utsumi T, Ueda T, Fukasawa S, et al. Prognostic models for renal cell carcinoma recurrence: External validation in a Japanese population. Int J Urol 2011;18:667-71.

(22.) Brookman-Amissah S, Kendel F, Spivak I, et al. Impact of clinical variables on predicting disease-free survival of patients with surgically resected renal cell carcinoma. BJU Int 2009;103:1375-80. http://dx.doi.org/10.1111/j.1464-410X.2008.08233.x

(23.) Klatte T, Patard JJ, de Martino M, et al. Tumor size does not predict risk of metastatic disease or prognosis of small renal cell carcinomas. J Urol 2008;179:1719-26. http://dx.doi.org/10.1016/j.juro.2008.01.018

(24.) Zastrow S, Brookman-May S, Cong TA, et al. Decision curve analysis and external validation of the postoperative Karakiewicz nomogram for renal cell carcinoma based on a large single-center study cohort. World J Urol 2015;33:381-8. http://dx.doi.org/10.1007/s00345-014-1321-6. Epub 2014 May 22.

(25.) Meskawi M, Sun M, Trinh QD, et al. A review of integrated staging systems for renal cell carcinoma. Eur Urol 2012;62:303-14. http://dx.doi.org/10.1016/j.eururo.2012.04.049

Correspondence: Dr. Giorgio Ivan Russo, Department of Urology, University of Catania, Catania, Italy; giorgioivan@virgilio.it

Giorgio Ivan Russo, MD; Alessandro Di Rosa, MD; Vincenzo Favilla, MD; Eugenia Fragala, MD; Tommaso Castelli, MD; Salvatore Privitera, MD; Sebastiano Cimino, MD; Giuseppe Morgia, MD

Department of Urology, University of Catania, Catania, Italy

Caption: Fig. 1. Flow diagram of included studies.

Caption: Fig. 2. Analysis of risk of publication bias. Funnel plot of studies included in meta-analysis on disease recurrence free survival (A) and cancer-specific survival (B). The effect of each study is marked by a circle. Uneven distributions of the studies around 95% confidence interval line should suggest the presence of publication bias, which is not the case in this funnel plot. SE: standard error; OR: odds ratio.

Table 1. Characteristics of the included studies

Reference               Data source            Model         No.
                                                           patients

Kattan et al.,       Single institution       Kattan         601
  2001 (9)                                    Cindolo
Liu et al.,          Single institution       Kattan         653
  2009 (12)                                Postoperative
                                            Karakiewicz
Cindolo et al.,      Single institution       Cindolo        660
  2003 (8)
Cindolo et al.,      Multi institution        Cindolo        2404
  2005 (4)                                    Kattan
Cindolo et al.,      Multi institution     Preoperative      3230
  2013 (1)                                  Karakiewicz
                                           Postoperative
                                            Karakiewicz
Karakiewicz et       Multi institution     Preoperative      1972
  al., 2009 (7)                             Karakiewicz
Karakiewicz et       Multi institution     Postoperative     3560
  al., 2009 (13)                            Karakiewicz
Karakiewicz et       Multi institution     Postoperative     2530
  al., 2007 (6)                             Karakiewicz
Kutikov et al.,      Multi institution     Preoperative      3560
  2010 (14)                                 Karakiewicz
Gontero et al.,      Multi institution     Preoperative      3364
  2013 (15)                                 Karakiewicz
                                              Kattan
Tan et al., 2011     Single institution    Postoperative     390
  (18)                                      Karakiewicz
Hupertan et al.,     Single institution       Kattan         565
  2006 (16)
Utsumi et al.,         Multi       CUH        Kattan         152
  2011 (19)         institution   Centre      Cindolo
                                   CCC        Kattan          65
                                  Centre      Cindolo
Suzuki et al.,       Multi institution        Kattan         211
  2011 (17)
Klatte et al.,       Multi institution        Kattan         995
  2008 (21)                                Postoperative
                                            Karakiewicz
Brookman-Amissah,    Single institution       Cindolo        771
  2009 (20)
Zastrow et al.,      Single institution    Postoperative     1480
  2013 (22)                                 Karakiewicz

Reference               Data source            No.       Follow-up
                                           recurrences   (median)

Kattan et al.,       Single institution        66           40
  2001 (9)
Liu et al.,          Single institution        156          65
  2009 (12)
Cindolo et al.,      Single institution        110          42
  2003 (8)
Cindolo et al.,      Multi institution         541          42
  2005 (4)
Cindolo et al.,      Multi institution         N/A          49
  2013 (1)
Karakiewicz et       Multi institution         N/A          42
  al., 2009 (7)
Karakiewicz et       Multi institution         N/A          32
  al., 2009 (13)
Karakiewicz et       Multi institution         N/A          39
  al., 2007 (6)
Kutikov et al.,      Multi institution         N/A         45.6
  2010 (14)
Gontero et al.,      Multi institution         N/A          48
  2013 (15)
Tan et al., 2011     Single institution        98           65
  (18)
Hupertan et al.,     Single institution        101          60
  2006 (16)
Utsumi et al.,         Multi       CUH         36           60
  2011 (19)         institution   Centre
                                   CCC          6           60
                                  Centre
Suzuki et al.,       Multi institution         41           81
  2011 (17)
Klatte et al.,       Multi institution         52           37
  2008 (21)
Brookman-Amissah,    Single institution        173          67
  2009 (20)
Zastrow et al.,      Single institution        N/A          82
  2013 (22)

Reference               Data source         Outcomes

Kattan et al.,       Single institution       DFS
  2001 (9)                                 OS-CSS-DFS
Liu et al.,          Single institution    OS-CSS-DFS
  2009 (12)                                OS-CSS-DFS
Cindolo et al.,      Single institution       DFS
  2003 (8)
Cindolo et al.,      Multi institution     OS-CSS-DFS
  2005 (4)                                 OS-CSS-DFS
Cindolo et al.,      Multi institution        CSS
  2013 (1)                                    CSS

Karakiewicz et       Multi institution        CSS
  al., 2009 (7)
Karakiewicz et       Multi institution        CSS
  al., 2009 (13)
Karakiewicz et       Multi institution        CSS
  al., 2007 (6)
Kutikov et al.,      Multi institution        CSS
  2010 (14)
Gontero et al.,      Multi institution        CSS
  2013 (15)                                OS-CSS-DFS
Tan et al., 2011     Single institution    OS-CSS-DFS
  (18)
Hupertan et al.,     Single institution     CSS-DFS
  2006 (16)
Utsumi et al.,         Multi       CUH        DFS
  2011 (19)         institution   Centre      DFS
                                   CCC        DFS
                                  Centre      DFS
Suzuki et al.,       Multi institution        DFS
  2011 (17)
Klatte et al.,       Multi institution      CSS-DFS
  2008 (21)                                 CSS-DFS

Brookman-Amissah,    Single institution       DFS
  2009 (20)
Zastrow et al.,      Single institution       CSS
  2013 (22)

Reference               Data source               AUC

Kattan et al.,       Single institution          0.740
  2001 (9)                                 0.700-0.715-0.752
Liu et al.,          Single institution    0.752-0.793-0.841
  2009 (12)                                0.716-0.754-0.785
Cindolo et al.,      Single institution           N/A
  2003 (8)
Cindolo et al.,      Multi institution     0.615-0.648-0.672
  2005 (4)                                 0.706-0.771-0.807
Cindolo et al.,      Multi institution           0.784
  2013 (1)                                       0.842
Karakiewicz et       Multi institution           0.842
  al., 2009 (7)
Karakiewicz et       Multi institution           0.867
  al., 2009 (13)
Karakiewicz et       Multi institution           0.865
  al., 2007 (6)
Kutikov et al.,      Multi institution           0.867
  2010 (14)
Gontero et al.,      Multi institution           0.860
  2013 (15)                                0.670-0.730-0.730

Tan et al., 2011     Single institution    0.770-0.840-0.810
  (18)
Hupertan et al.,     Single institution       0.847-0.607
  2006 (16)
Utsumi et al.,         Multi       CUH           0.795
  2011 (19)         institution   Centre         0.700
                                   CCC           0.745
                                  Centre         0.634
Suzuki et al.,       Multi institution           0.735
  2011 (17)
Klatte et al.,       Multi institution        0.778-0.755
  2008 (21)                                   0.724-0.704

Brookman-Amissah,    Single institution          0.690
  2009 (20)
Zastrow et al.,      Single institution          0.905
  2013 (22)

AUC: area under the curve; OS: overall survival; CSS:
cancer-specific survival; DFS: disease-free survival; CUH:
Chiba University Hospital; CCC: Chiba Cancer Center; N/A:
not applicable.

Table 2. Summary of the pooled ORs and corresponding AUCs of each
models for predictive capability of disease recurrence free survival

                     Postoperative      Kattan        Cindolo
                      Karakiewicz

No. studies                3              8              4
Heterogeneity test
  [x.sup.2]              0.19           4.012           3.53
  df                       2              8              4
  p value                0.91            0.86           0.94
  Combined ORs
  OR                     4.32            2.97           3.89
  95% CI              1.13-16.47      1.66-5.34      2.06-7.34
  Converted AUC      0.728 (0.01)    0.675 (0.01)   0.713 (0.01)
    (SE)
  Gain in                0.053          -0.038
    predictive        (<0.01) (a)     (0.03) (c)
    accuracy %           0.015
    (p value)         (0.432) (b)

OR: odds ratio; AUC: area under the curve; df: degree of freedom;
CI: confidential interval; SE: standard error. (a) Postoperative
Karakiewicz vs. Kattan; (b) Postoperative Karakiewicz vs. Cindolo;
(c) Kattan vs. Cindolo.

Table 3. Summary of the pooled ORs and corresponding AUCs of each
models for predictive capability of CSS

                           Preoperative        Postoperative
                            Karakiewicz         Karakiewicz

No. studies                      4                   7
Heterogeneity test
[chi square]                   0.40                0.40
df                               3                   6
p value                        0.94                1.00
Combined Odds Ratio
OR                             8.47                8.82
95%CI                       2.79-25.70          2.08-37.40
Converted AUC (SE)         0.810 (0.01)        0.814 (0.01)
Gain in predictive       0.030 (0.020) (a)   0.004 (0.730) (b)
  accuracy % (p value)   0.155 (<0.01) (c)   0.034 (<0.01) (d)
                                             0.159 (<0.01) (e)

                              Kattan           Cindolo

No. studies                      4                2
Heterogeneity test
[chi square]                   0.02              0.20
df                               3                1
p value                        1.00              0.59
Combined Odds Ratio
OR                             6.52              2.61
95%CI                       1.80-23.57        1.58-4.30
Converted AUC (SE)         0.780 (0.01)      0.655 (0.01)
Gain in predictive       0.125 (<0.01) (f)        --
  accuracy % (p value)

OR: odds ratio; AUC: area under the curve; CSS: cancer-specific
survival; df: degree of freedom; CI: confidential interval; SE:
standard error. (a) Preoperative Karakiewicz vs. Kattan;
(b) Preoperative Karakiewicz vs. Postoperative Karakiewicz;
(c) Preoperative Karakiewicz vs. Cindolo; (d) Postoperative
Karakiewicz vs. Kattan; (e) Postoperative Karakiewicz vs.
Cindolo; (f) Kattan vs. Cindolo.

Fig. 3. Forest plot for postoperative Karakiewicz, Kattan and
Cindolo nomograms in predicting disease recurrence-free survival.

Study or Subgroup        log[Odds Ratio]      Disease           Free
                                            Recurrence       Disease
                                                          Recurrence

                                             SE   Total        Total
2.1.2 Post operative Karakiewicz

Klatte 2008                          1.3   0.78      52          943
Liu 2009                            1.92   1.76     156          497
Tan 2011                            2.14   2.38      98          292
Subtotal (95% CI)                                   306         1732

Heterogeneity: [Chi.sup.2] = 0.19,
df = 2 (P = 0.91); [I.sup.2] = 0%

Test for overall effect: Z = 2.14 (P = 0.03)

2.1.3 Kattan

Cindolo 2005                        2.11   0.91     152         2252
Hupertan 2006                       0.64   0.38     101          464
Kattan 2001                         1.56   1.32      66          535
Klatte 2008                         1.67   1.25      52          943
Liu 2009                            2.45   2.53     156          497
Suzuki 2011                         1.52   3.96      41          170
Tan 2011                             1.5   1.05      98          292
Utsuml 2011 CCC cohort              1.59   4.11      36          116
Utsuml 2011 CUH cohort                 2   1.23       6           59
Subtotal (95% CI)                                   708         5328

Heterogeneity: [Chi.sup.2] = 4.02,
df = 8 (P = 0.86); [I.sup.2] = 0%

Test for overall effect: Z = 3.65 (P = 0.0003)

2.1.4 Cindolo

Brookman-Amissah 2009               1.19   0.55     173          598
Cindolo 2005                        1.07   0.48     152         2252
Liu 2009                            1.64   0.98     156          497
Utsuml 2011 CCC cohort              2.28   1.85      36          116
Utsumi 2011 CUH cohort              3.56   1.33       6           59
Subtotal (95% CI)                                   523         3522

Heterogeneity: [Chi.sup.2] = 3.53,
df = 4 (P = 0.47); [I.sup.2] = 0%

Test for overall effect: Z = 4.20 (P < 0.0001)

Total (95% CI)                                     1537        10582

Heterogeneity: [Chi.sup.2] = 8.23,
df = 16 (P = 0.94); [I.sup.2] = 0%

Test for overall effect: Z = 5.92 (P < 0.00001)

Test for subgroup differences:
[Chi.sup.2] = 0.49. df = 2 (P = 0.78).
[I.sup.2] = 0%

Study or Subgroup        Weight              Odds Ratio
                                             IV, Fixed,
                                                 95% CI

2.1.2 Post operative Karakiewicz

Klatte 2008                7.2%       3.67 [0.80,16.92]
Liu 2009                   1.4%     6.82 [0.22, 214.76]
Tan 2011                   0.8%     8.50 [0.08, 902.09]
Subtotal (95% CI)          9.4%      4.32 [1.13, 16.47]

Heterogeneity: [Chi.sup.2] = 0.19,
df = 2 (P = 0.91); [I.sup.2] = 0%

Test for overall effect: Z = 2.14 (P = 0.03)

2.1.3 Kattan

Cindolo 2005               5.3%      8.25 [1.39, 49.09]
Hupertan 2006             30.3%       1.90 [0.90, 3.99]
Kattan 2001                2.5%      4.76 [0.36, 63.25]
Klatte 2008                2.8%      5.31 [0.46, 61.56]
Liu 2009                   0.7%   11.59 [0.08, 1650.29]
Suzuki 2011                0.3%   4.57 [0.00, 10737.07]
Tan 2011                   4.0%      4.48 [0.57, 35.09]
Utsuml 2011 CCC cohort     0.3%   4.90 [0.00, 15451.35]
Utsuml 2011 CUH cohort     2.9%      7.39 [0.66, 82.33]
Subtotal (95% CI)         48.9%       2.97 [1.66, 5.34]

Heterogeneity: [Chi.sup.2] = 4.02,
df = 8 (P = 0.86); [I.sup.2] = 0%

Test for overall effect: Z = 3.65 (P = 0.0003)

2.1.4 Cindolo

Brookman-Amissah 2009     14.4%       3.29 [1.12, 9.66]
Cindolo 2005              19.0%       2.92 [1.14, 7.47]
Liu 2009                   4.6%      5.16 [0.76, 35.19]
Utsuml 2011 CCC cohort     1.3%     9.78 [0.26, 367.21]
Utsumi 2011 CUH cohort     2.5%    35.16 [2.59, 476.64]
Subtotal (95% CI)         41.7%        3.89 [2.06,7.34]

Heterogeneity: [Chi.sup.2] = 3.53,
df = 4 (P = 0.47); [I.sup.2] = 0%

Test for overall effect: Z = 4.20 (P < 0.0001)

Total (95% CI)           100.0%        3.44 [2.29,5.19]

Heterogeneity: [Chi.sup.2] = 8.23,
df = 16 (P = 0.94); [I.sup.2] = 0%

Test for overall effect: Z = 5.92 (P < 0.00001)

Test for subgroup differences:
[Chi.sup.2] = 0.49. df = 2 (P = 0.78).
[I.sup.2] = 0%

Fig. 4. Forest plot for preoperative Karakiewicz, postoperative
Karakiewicz, Kattan and Cindolo nomograms in predicting
cancer-specific survival.

Cancer-Specifical Survival.

Study or Subgroup    log [Odds    Cancer         Cancer
                       Ratio]     Specific       Specific
                                  Deaths         Survival

                                   SE    Total   Total   Weight

2.2.1 Pre-operative Karakiewicz Nomogram

Cindolo 2013             1.88     0.7     416    2753     9.0%
Gontero 2013             2.65     1.4     834    2530     2.3%
Karakiewicz 2009         2.45     1.87    272    1700     1.3%
Kutikov 2010             2.75 1.89        568    2992     1.2%
Subtotal (95% CI)                        2090    9975    13.8%

Heterogeneity: [Chi.sup.2] = 0.40, df= 3
(P = 0.94); [I.sup.2] = 0%

Test for overall effect: Z= 3.77 (P = 0.0002)

2.2.2 Post operative Karakiewicz Nomogram

Cindolo 2013             2.47     1.56    416    2483     1.8%
Karakiewicz 2007          2.7     1.76    598    1932     1.4%
Karakiewicz 2009-2       2.74     2.68    269    3291     0.6%
Klatte 2008              2.77     5.79     61     934     0.1%
Liu 2009                 1.65     1.09    123     560     3.7%
Tan 2011                 2.48     4.12     63     390     0.3%
Zastrow 2014             3.27     5.24    268    1480     0.2%
Subtotal (95% CI)                        1798    11070    8.1%

Heterogeneity: [Chi.sup.2] = 0.46, df = 6
(P = 1.00); [I.sup.2] = 0%

Test for overall effect: Z= 2.95 (P = 0.003)

2.2.4 Kattan Nomogram

Cindolo 2005             1.84     0.79    360    2044     7.1%
Klatte 2008              1.86     1.92     61     934     1.2%
Liu 2009                 1.96     1.56    123     530     1.8%
Tan 2011                 2.48     4.88     63     327     0.2%
Subtotal (95% CI)                         607    3835    10.3%

Heterogeneity: [Chi.sup.2]= 0.02, df= 3
(P = 1.00); [I.sup.2] = 0%

Test for overall effect: Z= 2.86 (P = 0.004)

2.2.5 Cindolo Nomogram

Cindolo 2005             0.91     0.27    360    2044    60.6%
Liu 2009                 1.36     0.78    123     530     7.3%
Subtotal (95% CI)                         483    2574    67.8%

Heterogeneity: [Chi.sup.2] = 0.30, df = 1
(P = 0.59); [I.sup.2] = 0%

Test for overall effect: Z = 3.76 (P = 0.0002)

Total (95% CI)                           4978    27454   100.0%

Heterogeneity: [Chi.sup.2]= 7.34,
df = 16 (P = 0.97); la= 0%

Test for overall effect: Z= 6.25 (P < 0.00001)

Test for subgroup differences: [Chi.sup.2]= 6.16. df= 3
(P = 0.10). [I.sup.2] = 51.3%

Study or Subgroup    Odds Ratio

                     IV, Fixed, 95% CI

2.2.1 Pre-operative Karakiewicz Nomogram

Cindolo 2013              6.55 [1.66, 25.84]
Gontero 2013             14.15 [0.91,220.07]
Karakiewicz 2009        11.59 [0.30, 452.66]
Kutikov 2010            15.64 [0.39, 635.45]
Subtotal (95% CI)          8.47 [2.79,25.70]

Heterogeneity: [Chi.sup.2] = 0.40, df= 3
(P = 0.94); [I.sup.2] = 0%

Test for overall effect: Z= 3.77 (P = 0.0002)

2.2.2 Post operative Karakiewicz Nomogram

Cindolo 2013            11.82 [0.56, 251.53]
Karakiewicz 2007        14.88 [0.47, 468.50]
Karakiewicz 2009-2     15.49 [0.08, 2959.29]
Klatte 2008          15.96 [0.00,1353482.52]
Liu 2009                   5.21 [0.61,44.10]
Tan 2011              11.94 [0.00, 38370.78]
Zastrow 2014         26.31 [0.00, 759344.83]
Subtotal (95% CI)         8.82 [2.08, 37.40]

Heterogeneity: [Chi.sup.2] = 0.46, df = 6
(P = 1.00); [I.sup.2] = 0%

Test for overall effect: Z= 2.95 (P = 0.003)

2.2.4 Kattan Nomogram

Cindolo 2005              6.30 [1.34, 29.62]
Klatte 2008              6.42 [0.15. 276.75]
Liu 2009                  7.10 [0.33,151.04]
Tan 2011              11.94 [0.00,170182.09]
Subtotal (95% CI)         6.52 [1.80, 23.57]

Heterogeneity: [Chi.sup.2]= 0.02, df= 3
(P = 1.00); [I.sup.2] = 0%

Test for overall effect: Z= 2.86 (P = 0.004)

2.2.5 Cindolo Nomogram

Cindolo 2005               2.48 [1.46, 4.22]
Liu 2009                   3.90 [0.84,17.97]
Subtotal (95% CI)           2.61 [1.58,4.30]

Heterogeneity: [Chi.sup.2] = 0.30, df = 1
(P = 0.59); [I.sup.2] = 0%

Test for overall effect: Z = 3.76 (P = 0.0002)

Total (95% CI)             3.72 [2.46, 5.61]

Heterogeneity: [Chi.sup.2]= 7.34,
df = 16 (P = 0.97); la= 0%

Test for overall effect: Z= 6.25 (P < 0.00001)

Test for subgroup differences: [Chi.sup.2]= 6.16. df= 3
(P = 0.10). [I.sup.2] = 51.3%


@@@2557-10804-1-PB.pdf

----------

Please note: Illustration(s) are not available due to copyright restrictions.
COPYRIGHT 2015 Canadian Urological Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2015 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:ORIGINAL RESEARCH
Author:Russo, Giorgio Ivan; Di Rosa, Alessandro; Favilla, Vincenzo; Fragala, Eugenia; Castelli, Tommaso; Pr
Publication:Canadian Urological Association Journal (CUAJ)
Article Type:Report
Date:Jun 1, 2015
Words:6126
Previous Article:Fournier's gangrene: a retrospective analysis of 26 cases in a Canadian hospital and literature review.
Next Article:Artificial urinary sphincter erosion after radical prostatectomy in patients treated with and without radiation.
Topics:

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |