A1 Vertaisarvioitu alkuperäisartikkeli tieteellisessä lehdessä
A three-feature prediction model for metastasis-free survival after surgery of localized clear cell renal cell carcinoma
Tekijät: Mattila Kalle E., Laajala Teemu D., Tornberg Sara V., Kilpeläinen Tuomas P., Vainio Paula, Ettala Otto, Boström Peter J., Nisen Harry, Elo Laura L., Jaakkola Panu M.
Kustantaja: NATURE RESEARCH
Julkaisuvuosi: 2021
Journal: Scientific Reports
Tietokannassa oleva lehden nimi: SCIENTIFIC REPORTS
Lehden akronyymi: SCI REP-UK
Artikkelin numero: ARTN 8650
Vuosikerta: 11
Sivujen määrä: 8
ISSN: 2045-2322
eISSN: 2045-2322
DOI: https://doi.org/10.1038/s41598-021-88177-9
Rinnakkaistallenteen osoite: https://research.utu.fi/converis/portal/detail/Publication/56208582
After surgery of localized renal cell carcinoma, over 20% of the patients will develop distant metastases. Our aim was to develop an easy-to-use prognostic model for predicting metastasis-free survival after radical or partial nephrectomy of localized clear cell RCC. Model training was performed on 196 patients. Right-censored metastasis-free survival was analysed using LASSO-regularized Cox regression, which identified three key prediction features. The model was validated in an external cohort of 714 patients. 55 (28%) and 134 (19%) patients developed distant metastases during the median postoperative follow-up of 6.3 years (interquartile range 3.4-8.6) and 5.4 years (4.0-7.6) in the training and validation cohort, respectively. Patients were stratified into clinically meaningful risk categories using only three features: tumor size, tumor grade and microvascular invasion, and a representative nomogram and a visual prediction surface were constructed using these features in Cox proportional hazards model. Concordance indices in the training and validation cohorts were 0.755 +/- 0.029 and 0.836 +/- 0.015 for our novel model, which were comparable to the C-indices of the original Leibovich prediction model (0.734 +/- 0.035 and 0.848 +/- 0.017, respectively). Thus, the presented model retains high accuracy while requiring only three features that are routinely collected and widely available.
Ladattava julkaisu This is an electronic reprint of the original article. |  |
