Limits...
External validation of extended prostate biopsy nomogram.

Hrbáček J, Minárik I, Sieger T, Babjuk M - Cent European J Urol (2015)

Bottom Line: Area under the ROC curve calculated for our data using the nomogram was 0.773, similar to that reported originally.The nomogram yielded overall good predictive accuracy as measured by the area under the ROC curve, but it systematically overestimated PC probability in individual patients.However, we showed how the nomogram could easily be adapted to our patient sample, resolving the bias issue.

View Article: PubMed Central - PubMed

Affiliation: Department of Urology, 2 Faculty of Medicine and Motol University Hospital, Charles University Praha, Czech Republic.

ABSTRACT

Introduction: Historical nomograms for the prediction of cancer on prostate biopsy, developed in the sextant biopsy era are no more accurate today. The aim of this study was an independent external validation of a 10-core biopsy nomogram by Chun et al. (2007).

Material and methods: A total of 322 patients who presented for their initial biopsy in a tertiary care center and had all the necessary data available were included in the retrospective analysis. To validate the nomogram, receiver operator characteristic (ROC) curves and calibration plots were constructed.

Results: Area under the ROC curve calculated for our data using the nomogram was 0.773, similar to that reported originally. However, the nomogram systematically overestimated prostate cancer risk, which, for our data, could be resolved by subtracting 24 points from the total number of points of the nomogram.

Conclusions: The nomogram yielded overall good predictive accuracy as measured by the area under the ROC curve, but it systematically overestimated PC probability in individual patients. However, we showed how the nomogram could easily be adapted to our patient sample, resolving the bias issue.

No MeSH data available.


Related in: MedlinePlus

Predicted and observed probability of prostate cancer in our patient sample using the graphical tool from the original article by Chun et al. Point estimates of the probability of prostate cancer are given as triangles and supplemented with 95% confidence intervals based on the binomial distribution (vertical lines).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4526620&req=5

Figure 0002: Predicted and observed probability of prostate cancer in our patient sample using the graphical tool from the original article by Chun et al. Point estimates of the probability of prostate cancer are given as triangles and supplemented with 95% confidence intervals based on the binomial distribution (vertical lines).

Mentions: Calibration plots computed using the numerical formula and the graphical nomogram are shown in Figure 1A and 1B, respectively. An intriguing finding of our analysis was that the graphical nomogram (Figure 1A) systematically overestimated the prostate cancer risk. However, we could adapt the graphical nomogram to yield correct PC risk estimates in our patients. We did so by subtracting the value of 24 points from the total points obtained from the graphical nomogram. (This estimate was based on our data set and calculated using a logistic regression model of PC, in which all the parameters, but the absolute term were fixed to the parameters of the graphical nomogram.) The numerical nomogram (Figure 1B) yielded unbiased probabilities of PC.


External validation of extended prostate biopsy nomogram.

Hrbáček J, Minárik I, Sieger T, Babjuk M - Cent European J Urol (2015)

Predicted and observed probability of prostate cancer in our patient sample using the graphical tool from the original article by Chun et al. Point estimates of the probability of prostate cancer are given as triangles and supplemented with 95% confidence intervals based on the binomial distribution (vertical lines).
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4526620&req=5

Figure 0002: Predicted and observed probability of prostate cancer in our patient sample using the graphical tool from the original article by Chun et al. Point estimates of the probability of prostate cancer are given as triangles and supplemented with 95% confidence intervals based on the binomial distribution (vertical lines).
Mentions: Calibration plots computed using the numerical formula and the graphical nomogram are shown in Figure 1A and 1B, respectively. An intriguing finding of our analysis was that the graphical nomogram (Figure 1A) systematically overestimated the prostate cancer risk. However, we could adapt the graphical nomogram to yield correct PC risk estimates in our patients. We did so by subtracting the value of 24 points from the total points obtained from the graphical nomogram. (This estimate was based on our data set and calculated using a logistic regression model of PC, in which all the parameters, but the absolute term were fixed to the parameters of the graphical nomogram.) The numerical nomogram (Figure 1B) yielded unbiased probabilities of PC.

Bottom Line: Area under the ROC curve calculated for our data using the nomogram was 0.773, similar to that reported originally.The nomogram yielded overall good predictive accuracy as measured by the area under the ROC curve, but it systematically overestimated PC probability in individual patients.However, we showed how the nomogram could easily be adapted to our patient sample, resolving the bias issue.

View Article: PubMed Central - PubMed

Affiliation: Department of Urology, 2 Faculty of Medicine and Motol University Hospital, Charles University Praha, Czech Republic.

ABSTRACT

Introduction: Historical nomograms for the prediction of cancer on prostate biopsy, developed in the sextant biopsy era are no more accurate today. The aim of this study was an independent external validation of a 10-core biopsy nomogram by Chun et al. (2007).

Material and methods: A total of 322 patients who presented for their initial biopsy in a tertiary care center and had all the necessary data available were included in the retrospective analysis. To validate the nomogram, receiver operator characteristic (ROC) curves and calibration plots were constructed.

Results: Area under the ROC curve calculated for our data using the nomogram was 0.773, similar to that reported originally. However, the nomogram systematically overestimated prostate cancer risk, which, for our data, could be resolved by subtracting 24 points from the total number of points of the nomogram.

Conclusions: The nomogram yielded overall good predictive accuracy as measured by the area under the ROC curve, but it systematically overestimated PC probability in individual patients. However, we showed how the nomogram could easily be adapted to our patient sample, resolving the bias issue.

No MeSH data available.


Related in: MedlinePlus