Division of Nephrology, Department of Medicine, Durham, NC; Duke Clinical Research Institute, Duke University School of Medicine, Durham, NC.
Duke Clinical Research Institute, Duke University School of Medicine, Durham, NC.
Am J Kidney Dis. 2019 Dec;74(6):771-781. doi: 10.1053/j.ajkd.2019.05.026. Epub 2019 Aug 21.
RATIONALE & OBJECTIVE: An elevated fibroblast growth factor 23 (FGF-23) level is independently associated with adverse outcomes in populations with chronic kidney disease, but it is unknown whether FGF-23 testing can improve clinical risk prediction in individuals.
Prospective cohort study.
SETTING & PARTICIPANTS: Participants in the Chronic Renal Insufficiency Cohort (CRIC) Study (n = 3,789).
Baseline carboxy-terminal FGF-23 (cFGF-23) level.
All-cause and cardiovascular (CV) mortality, incident end-stage renal disease (ESRD), heart failure (HF) admission, and atherosclerotic events at 3, 5, and 8 years.
We assessed changes in model performance by change in area under the receiver operating characteristic curve (ΔAUC), integrated discrimination improvement (IDI), relative IDI, and net reclassification index (NRI) above standard clinical factors. We performed sensitivity analyses, including an additional model comparing the addition of phosphate rather than cFGF-23 level and repeating our analyses using an internal cross-validation cohort.
Addition of a single baseline value of cFGF-23 to a base prediction model improved prediction of all-cause mortality (ΔAUC, 0.017 [95% CI, 0.001-0.033]; IDI, 0.021 [95% CI, 0.006-0.036]; relative IDI, 32.7% [95% CI, 8.5%-56.9%]), and HF admission (ΔAUC, 0.008 [95% CI, 0.0004-0.016]; IDI, 0.019 [95% CI, 0.004-0.034]; relative IDI, 10.0% [95% CI, 1.8%-18.3%]), but not CV mortality, ESRD, or atherosclerotic events at 3 years of follow-up. The NRI did not reach statistical significance for any of the 3-year outcomes. The incremental predictive utility of cFGF-23 level diminished in analyses of the 5- and 8-year outcomes. The cFGF-23 models outperformed the phosphate model for each outcome.
Power to detect increased CV mortality likely limited by low event rate. The NRI is not generalizable without accepted prespecified risk thresholds.
Among individuals with CKD, single measurements of cFGF-23 improve prediction of risks for all-cause mortality and HF admission but not CV mortality, ESRD, or atherosclerotic events. Future studies should evaluate the predictive utility of repeated cFGF-23 testing.
成纤维细胞生长因子 23（FGF-23）水平升高与慢性肾脏病患者的不良预后独立相关，但尚不清楚 FGF-23 检测是否能改善个体的临床风险预测。
前瞻性队列研究。
慢性肾功能不全队列研究（CRIC）中的参与者（n=3789）。
基线羧基末端 FGF-23（cFGF-23）水平。
全因死亡率和心血管死亡率、新发终末期肾病（ESRD）、心力衰竭（HF）入院以及 3、5 和 8 年时的动脉粥样硬化事件。
通过接受者操作特征曲线下面积（AUC）的改变、综合鉴别改善（IDI）、相对 IDI 和净重新分类指数（NRI）来评估模型性能的变化，标准临床因素之外。我们进行了敏感性分析，包括比较添加磷酸盐而不是 cFGF-23 水平的额外模型，并使用内部交叉验证队列重复我们的分析。
在基础预测模型中添加单个基线 cFGF-23 值可改善全因死亡率的预测（AUC 的变化，0.017[95%置信区间，0.001-0.033]；IDI，0.021[95%置信区间，0.006-0.036]；相对 IDI，32.7%[95%置信区间，8.5%-56.9%]）和 HF 入院（AUC 的变化，0.008[95%置信区间，0.0004-0.016]；IDI，0.019[95%置信区间，0.004-0.034]；相对 IDI，10.0%[95%置信区间，1.8%-18.3%]），但不能改善心血管死亡率、ESRD 或 3 年时的动脉粥样硬化事件。在 3 年的随访中，NRI 没有达到统计学意义。cFGF-23 水平的增量预测效用在 5 年和 8 年的结果分析中减弱。对于每个结局，cFGF-23 模型的预测效果均优于磷酸盐模型。
检测心血管死亡率增加的能力可能受到低事件率的限制。没有接受预先指定的风险阈值，NRI 不具有普遍性。
在患有 CKD 的个体中，cFGF-23 的单次测量可改善全因死亡率和 HF 入院的风险预测，但不能改善心血管死亡率、ESRD 或动脉粥样硬化事件。未来的研究应评估重复 cFGF-23 检测的预测效用。
