Scholarlink

Background <p>In the era of risk-based prostate cancer (PCa) screening, overcoming the limitations of prostate-specific antigen (PSA) testing and stratifying men by individual risk is crucial. Our study aims to integrate anamnestic and lifestyle data with circulating biomarkers to minimize unnecessary second-level investigations (SLIs) for patients with suspected PCa, while improving the detection of clinically significant PCa (ISUP &gt; 1).</p> Methods <p>We collected plasma samples, recent clinical history, family cancer history, PSA levels, and lifestyle information from 904 men: 421 undergoing PSA testing, 421 with suspected and 62 with confirmed PCa. Univariable logistic regression was applied to identify ananmestic and lifestyle variables mostly associated with PCa. Penalized logistic regression models predictive of PCa or ISUP &gt; 1 PCa were built both using the 814 subjects with complete information for such variables, applying a 10-fold cross validation approach, and dividing the dataset into a training (<i>n</i> = 445: 132 PCa, 313 non-PCa) and a test (<i>n</i> = 369: 147 PCa, 222 non-PCa) set. The concentration of 50 sphingolipids was analysed on the latter set of 369 subjects by mass-spectrometry, and multivariable penalized regression with 10-fold cross-validation was applied to integrate anamnestic, lifestyle, sphingolipid data. ROC-AUCs on the test sets were compared with PSA ROC-AUCs.</p> Results <p>Age, cardiovascular disease (CVD), number of medications, and sedentariness were significantly associated with PCa detection and their combination with PSA improved its performance (ROC-AUC from 0.85 to 0.89). In the SLI subgroup (<i>n</i> = 437), adding age improved PSA predictive power (ROC-AUC from 0.60 to 0.70), but performance was still poor. Penalized regression with 10-fold cross-validation on the sphingolipid dataset identified hypertension, CVD, PSA, age, and five sphingolipids (HexCer-20, Cer-20, HexCer-24.1, GM3-24.1, DHCer-24) as key variables for accurate PCa classification (average ROC-AUC: 0.92). Cer-20 and CVD were consistently selected by models predicting ISUP &gt; 1 PCa. In the SLI subgroup, PSA, age, CVD, SM-16, HexCer-20, HexCer-24.1, DHS1P, and DHCer-24 were selected in all 10 models (average ROC-AUC: 0.83).</p> Conclusions <p>Circulating sphingolipids are promising biomarkers that, when combined with PSA, anamnestic, and lifestyle data, may enhance PCa screening precision and reduce the need for invasive, costly examinations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrating anamnestic and lifestyle data with sphingolipid levels for risk-based prostate cancer screening

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月14日

查看全文
  • Caterina Peraldo-Neia,
  • Paola Ostano,
  • Melissa Savioli,
  • Maurizia Mello-Grand,
  • Ilaria Gregnanin,
  • Francesca Guana,
  • Francesca Crivelli,
  • Francesco Montagnani,
  • Michele Dei-Cas,
  • Rita Paroni,
  • Antonella Sinopoli,
  • Francesco Ferranti,
  • Nicolò Testino,
  • Marco Oderda,
  • Andrea Zitella,
  • Chiara Fiameni,
  • Amedeo Gagliardi,
  • Alessio Naccarati,
  • Luca Clivio,
  • Paolo Gontero,
  • Stefano Zaramella,
  • Giovanna Chiorino

摘要

Background

In the era of risk-based prostate cancer (PCa) screening, overcoming the limitations of prostate-specific antigen (PSA) testing and stratifying men by individual risk is crucial. Our study aims to integrate anamnestic and lifestyle data with circulating biomarkers to minimize unnecessary second-level investigations (SLIs) for patients with suspected PCa, while improving the detection of clinically significant PCa (ISUP > 1).

Methods

We collected plasma samples, recent clinical history, family cancer history, PSA levels, and lifestyle information from 904 men: 421 undergoing PSA testing, 421 with suspected and 62 with confirmed PCa. Univariable logistic regression was applied to identify ananmestic and lifestyle variables mostly associated with PCa. Penalized logistic regression models predictive of PCa or ISUP > 1 PCa were built both using the 814 subjects with complete information for such variables, applying a 10-fold cross validation approach, and dividing the dataset into a training (n = 445: 132 PCa, 313 non-PCa) and a test (n = 369: 147 PCa, 222 non-PCa) set. The concentration of 50 sphingolipids was analysed on the latter set of 369 subjects by mass-spectrometry, and multivariable penalized regression with 10-fold cross-validation was applied to integrate anamnestic, lifestyle, sphingolipid data. ROC-AUCs on the test sets were compared with PSA ROC-AUCs.

Results

Age, cardiovascular disease (CVD), number of medications, and sedentariness were significantly associated with PCa detection and their combination with PSA improved its performance (ROC-AUC from 0.85 to 0.89). In the SLI subgroup (n = 437), adding age improved PSA predictive power (ROC-AUC from 0.60 to 0.70), but performance was still poor. Penalized regression with 10-fold cross-validation on the sphingolipid dataset identified hypertension, CVD, PSA, age, and five sphingolipids (HexCer-20, Cer-20, HexCer-24.1, GM3-24.1, DHCer-24) as key variables for accurate PCa classification (average ROC-AUC: 0.92). Cer-20 and CVD were consistently selected by models predicting ISUP > 1 PCa. In the SLI subgroup, PSA, age, CVD, SM-16, HexCer-20, HexCer-24.1, DHS1P, and DHCer-24 were selected in all 10 models (average ROC-AUC: 0.83).

Conclusions

Circulating sphingolipids are promising biomarkers that, when combined with PSA, anamnestic, and lifestyle data, may enhance PCa screening precision and reduce the need for invasive, costly examinations.