Modified GEE and Goodness of the Marginal Fit (GOMF) Test with Correlated Binary Responses for Contingency Tables

Association of Renal Stress/Damage and Filtration Biomarkers with Subsequent AKI during Hospitalization among Patients Presenting to the Emergency Department

BACKGROUND AND OBJECTIVES

Emergency departments (EDs) have a growing role in hospital admissions, but few studies address AKI biomarkers in the ED.

DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS

Patients admitted to the internal medicine service were enrolled during initial workup in the ED at Robert-Bosch-Hospital, Stuttgart, Germany. Daily serum creatinine (sCr) and urine output (UO) were recorded for AKI classification by Kidney Disease Improving Global Outcomes (KDIGO) criteria. Cystatin C, kidney injury molecule-1, liver-type fatty acid-binding protein, and neutrophil gelatinase-associated lipocalin were measured in blood and urine, and IL-18, insulin-like growth factor-binding protein 7 (IGFBP7), tissue inhibitor of metalloproteinases-2 (TIMP-2) and [TIMP-2]⋅[IGFBP7] were measured in urine collected at enrollment, after 6 hours, and the following morning. Association between these biomarkers and the end point of moderate-severe AKI (KDIGO stage 2-3) occurring within 12 hours of each sample collection was examined using generalized estimating equation logistic regression. Performance for prediction of the AKI end point using two previously validated [TIMP-2]-[IGFBP7] cutoffs was also tested.

RESULTS

Of 400 enrolled patients, 298 had sufficient sCr and UO data for classification by KDIGO AKI criteria: AKI stage 2 developed in 37 patients and AKI stage 3 in nine patients. All urinary biomarkers, sCr, and plasma cystatin C had statistically significant (P<0.05) odds ratios (ORs) for the AKI end point. In a multivariable model of the urine biomarkers and sCr, only [TIMP-2]⋅[IGFBP7] and sCr had statistically significant ORs. Compared with [TIMP-2]⋅[IGFBP7]<0.3 (ng/ml)(2)/1000, values between 0.3 and 2.0 (ng/ml)(2)/1000 indicated 2.5 (95% confidence interval [95% CI], 1.1 to 5.2) times the odds for the AKI end point and values >2.0 (ng/ml)(2)/1000 indicated 11.0 (95% CI, 4.4 to 26.9) times the odds. Addition of [TIMP-2]⋅[IGFBP7] to a clinical model significantly improved area under the receiver-operating characteristic curve from 0.67 (95% CI, 0.61 to 0.78) to 0.77 (95% CI, 0.72 to 0.86) (P<0.001); however, including both markers in the model was not significantly different from including either marker alone.

CONCLUSIONS

Urinary [TIMP-2]⋅[IGFBP7] with pre-established cutoffs provides valuable information about risk for imminent AKI in the ED that is complementary to sCr and clinical risk factors.

A nonparametric smoothing method for assessing GEE models with longitudinal binary data

Studies involving longitudinal binary responses are widely applied in the health and biomedical sciences research and frequently analyzed by generalized estimating equations (GEE) method. This article proposes an alternative goodness-of-fit test based on the nonparametric smoothing approach for assessing the adequacy of GEE fitted models, which can be regarded as an extension of the goodness-of-fit test of le Cessie and van Houwelingen (Biometrics 1991; 47:1267-1282). The expectation and approximate variance of the proposed test statistic are derived. The asymptotic distribution of the proposed test statistic in terms of a scaled chi-squared distribution and the power performance of the proposed test are discussed by simulation studies. The testing procedure is demonstrated by two real data.