ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005

email@asn-online.org

202-640-4660

The Latest on Twitter

Kidney Week

Abstract: SA-PO489

Prediction of Autosomal Dominant Polycystic Kidney Disease (ADPKD) Outcomes at Birth

Session Information

  • ADPKD: Clinical Studies
    October 27, 2018 | Location: Exhibit Hall, San Diego Convention Center
    Abstract Time: 10:00 AM - 12:00 PM

Category: Genetic Diseases of the Kidney

  • 1001 Genetic Diseases of the Kidney: Cystic

Authors

  • Mrug, Michal, University of Alabama at Birmingham, Birmingham, Alabama, United States
  • Shi, Tiange, University of Pittsburgh, Pittsburgh, Pennsylvania, United States
  • Torres, Vicente E., Mayo Clinic, Rochester, Minnesota, United States
  • Rahbari-Oskoui, Frederic F., Emory University School of Medicine, Atlanta, Georgia, United States
  • Bae, Kyongtae Ty, University of Pittsburgh, Pittsburgh, Pennsylvania, United States
  • Harris, Peter C., Mayo Clinic, Rochester, Minnesota, United States
  • Guay-Woodford, Lisa M., Children's National Health System, Washington, District of Columbia, United States
  • Bennett, William M., Legacy Good Samaritan Medical Center, Portland, Oregon, United States
  • Yu, Alan S.L., University of Kansas Medical Center, Kansas City, Kansas, United States
  • Chapman, Arlene B., University of Chicago, Chicago, Illinois, United States
Background

The identification of high-risk ADPKD patients at birth may help them to receive appropriate medical attention.

Methods

We developed logistic, tree and random forests models for reaching CKD stage 3b progression endpoint (GFR<45 ml/min/1.73m2) based on data collected by the Consortium for Radiologic Imaging Studies of Polycystic Kidney Disease study (CRISP; 241 ADPKD adults). We validated these models using data collected by the HALT-A study (558 ADPKD adults). In addition to the GFR endpoint and age, analyzed variables included race, gender, birth weight, gene type and mutation strength-based indices (data available at birth).

Results

The logistic regression model yielded area under an ROC curve (AUC) 0.786 (95% CI: 0.7231, 0.8491; p<0.001). The tree model pruned to the optimal complexity parameter of 0.038 had a similar prognostic ability as the logistic model (their 95% CI of AUC overlapped). Random forests grown by 1000 bootstrap replicate samples of a size of 196 samples with replacement improved the prognostic ability of the factors available at birth and reached AUC 0.91 (95% CI:0.8689, 0.9467; p<0.001). Validation of these models in HALT-A cohort revealed much lower AUC for the logistic regression model (AUC 0.537), the tree model (AUC 0.552), as well as random forests (AUC 0.612). This likely occurred in part due to overfitting during analyses of CRISP data.

Conclusion

The data available at birth have modest value in prediction of advanced CKD outcomes based on information collected in the CRISP and HALT-A studies. Our analyses also demonstrate an essential role of predictive model validation using independent cohort datasets.

Funding

  • NIDDK Support