Prediction of prostate cancer in unscreened men: External validation of a risk calculator

详细信息    查看全文

• 作者：Heidi A. van Vugt^a ; ^b ; ^{h.a.vanvugt@erasmusmc.nl} ; Monique J. Roobol^a ; Ries Kranse^c ; Liisa M&#xe4 ; &#xe4 ; tt&#xe4 ; nen^d ; Patrik Finne^e ; Jonas Hugosson^f ; Chris H. Bangma^a ; Fritz H. Schrö ; der^g ; Ewout W. Steyerberg^b
• 关键词：Nomogram ; Prostate biopsy ; Prostate cancer ; Prostate cancer risk calculator ; Risk stratification ; Screening ; Validation
• 刊名：European Journal of Cancer
• 出版年：2011
• 出版时间：April 2011
• 年：2011
• 卷：47
• 期：6
• 页码：903-909
• 全文大小：879 K

文摘

Background

Prediction models need external validation to assess their value beyond the setting where the model was derived from.

Objective

To assess the external validity of the European Randomized study of Screening for Prostate Cancer (ERSPC) risk calculator (www.prostatecancer-riskcalculator.com) for the probability of having a positive prostate biopsy (P(posb)).

Design, setting and participants

The ERSPC risk calculator was based on data of the initial screening round of the ERSPC section Rotterdam and validated in 1825 and 531 men biopsied at the initial screening round in the Finnish and Swedish sections of the ERSPC respectively. P(posb) was calculated using serum prostate specific antigen (PSA), outcome of digital rectal examination (DRE), transrectal ultrasound and ultrasound assessed prostate volume.

Measurements

The external validity was assessed for the presence of cancer at biopsy by calibration (agreement between observed and predicted outcomes), discrimination (separation of those with and without cancer), and decision curves (for clinical usefulness).

Results and limitations

Prostate cancer was detected in 469 men (26 % ) of the Finnish cohort and in 124 men (23 % ) of the Swedish cohort. Systematic miscalibration was present in both cohorts (mean predicted probability 34 % versus 26 % observed, and 29 % versus 23 % observed, both p < 0.001). The areas under the curves were 0.76 and 0.78, and substantially lower for the model with PSA only (0.64 and 0.68 respectively). The model proved clinically useful for any decision threshold compared with a model with PSA only, PSA and DRE, or biopsying all men. A limitation is that the model is based on sextant biopsies results.

Conclusions

The ERSPC risk calculator discriminated well between those with and without prostate cancer among initially screened men, but overestimated the risk of a positive biopsy. Further research is necessary to assess the performance and applicability of the ERSPC risk calculator when a clinical setting is considered rather than a screening setting.