Covariate-specific and time-dependent area-under-curve (AUC) is often used to evaluate the discriminative performance of biomarkers with time-to-event outcomes, particularly when certain covariates influence biomarkers’ accuracy. In biomarker research, despite extensive efforts, missing data remain unavoidable, with nonignorable missingness posing significant challenges. This article focuses on estimating the impact of covariates on time-dependent AUC in the presence of nonignorable missing biomarkers. Assuming a parametric model on the missing probability, we leverage instrumental variables to address the identifiable issue and estimate the unknown parameters. We integrate the inverse probability weighting approach into the score equation of a pseudo partial likelihood for estimation and inference. Additionally, we establish the asymptotic properties of the proposed estimators. Through simulation studies, we evaluate finite sample performance of the proposed estimators, and apply the method to data from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database.
OldenhuisCOostingSGietemaJet al.Prognostic versus predictive value of biomarkers in oncology. Eur J Cancer2008; 44: 946–953.
2.
FDA-NIHBiomarker Working Group. Prognostic biomarker. In BEST (biomarkers, endpointS, and other tools) resource [Internet]. Silver Spring, MD: Food and Drug Administration (US), 2016.
3.
HeagertyPJZhengY. Survival model predictive accuracy and ROC curves. Biometrics2005; 61: 92–105.
4.
PepeMSEtzioniRFengZet al.Phases of biomarker development for early detection of cancer. J Natl Cancer Inst2001; 93: 1054–1061.
5.
DoddLEPepeMS. Semiparametric regression for the area under the receiver operating characteristic curve. J Am Stat Assoc2003; 98: 409–417.
6.
JanesHPepeMS. Adjusting for covariates in studies of diagnostic, screening, or prognostic markers: an old concept in a new setting. Am J Epidemiol2008; 168: 89–97.
7.
HungHChiangCT. Estimation methods for time-dependent AUC models with survival data. Can J Stat2010; 38: 8–26.
8.
JiangXLiWLiRet al.Addressing subject heterogeneity in time-dependent discrimination for biomarker evaluation. Stat Med2024; 43: 1341–1353.
9.
JiangXLiWWangKet al.Analyzing heterogeneity in biomarker discriminative performance through partial time-dependent receiver operating characteristic curve modeling. Stat Methods Med Res2024; 33: 1424–1436.
10.
AnoopASinghPKJacobRSet al.CSF biomarkers for Alzheimer’s disease diagnosis. Int J Alzheimers Dis2010; 2010: 606802.
11.
BuéeLBussièreTBuée-ScherrerVet al.Tau protein isoforms, phosphorylation and role in neurodegenerative disorders. Brain Res Brain Res Rev2000; 33: 95–130.
12.
BlennowKVanmechelenEHampelH. CSF total tau, A42 and phosphorylated tau protein as biomarkers for Alzheimer’s disease. Mol Neurobiol2001; 24: 87–97.
13.
SunderlandTLinkerGMirzaNet al.Decreased -Amyloid1-42 and increased tau levels in cerebrospinal fluid of patients with Alzheimer disease. JAMA2003; 289: 2094–2103.
14.
LittleRJARubinDB. Theory of inference based on the likelihood function. In Statistical analysis with missing data. John Wiley & Sons, Ltd., 2002, pp.95–132. DOI: 10.1002/9781119013563.ch6.
15.
KimJKYuCL. A semiparametric estimation of mean functional with nonignorable missing data. J Am Stat Assoc2011; 106: 157–165.
16.
GelfandAESahuSK. Identifiability, improper priors, and Gibbs sampling for generalized linear models. J Am Stat Assoc1999; 94: 247–253.
17.
WangSShaoJKimJK. An instrumental variable approach for identification and estimation with nonignorable nonresponse. Stat Sin2014; 24: 1097–1116.
18.
RiddlesMKKimJKImJ. A propensity-score-adjustment method for nonignorable nonresponse. J Surv Stat Methodol2016; 4: 215–245.
19.
TsiatisAA. Semiparametric theory and missing data. New York, NY: Springer, 2006.
20.
TangNZhaoPZhuH. Empirical likelihood for estimating equations in nonignorable missing data. Stat Sin2014; 24: 723–747.
21.
ZhaoJShaoJ. Semiparametric pseudo-likelihood in generalized linear models with nonignorable missing data. J Am Stat Assoc2015; 110: 1577–1590.
22.
LiSNingY. Estimation of covariate-specific time-dependent ROC curves in the presence of missing biomarkers. Biometrics2015; 71: 666–676.
23.
ChengWLiX. A semi-parametric approach for time-dependent ROC curves with nonignorable missing biomarker. J Biopharm Stat2023; 33: 555–574.
CaiTDoddLE. Regression analysis for the partial area under the ROC curve. Stat Sin2008; 18: 817–836.
26.
NeweyWKMcFaddenD. Chapter 36 Large sample estimation and hypothesis testing. In Handbook of econometrics. Vol. 4. Elsevier, 1994. pp.2111–2245. DOI: 10.1016/S1573-4412(05)80005-4.
27.
SunYMcCullochCEMarrKAet al.Recurrent events analysis with data collected at informative clinical visits in electronic health records. J Am Stat Assoc2021; 116: 594–604.
28.
HuTNanBLinX. Proportional cross-ratio model. Lifetime Data Anal2019; 25: 480–506.
29.
KimSSchaubelDEMcCulloughKP. A C-index for recurrent event data: application to hospitalizations among dialysis patients. Biometrics2018; 74: 734–743.
30.
RaulinACDossSVTrottierZAet al.ApoE in Alzheimer’s disease: pathophysiology and therapeutic strategies. Mol Neurodegener2022; 17: 72.
31.
BatishSHaji-SheikhiFKareAet al.Cerebrospinal fluid ab42, total tau, and phosphorylated tau in the evaluation of Alzheimer’s dementia: experience from a commercial reference laboratory. Alzheimers Dement2021; 17: e054422.
32.
PettigrewCSoldanAWangJet al.Longitudinal CSF Alzheimer’s disease biomarker changes from middle age to late adulthood. Alzheimers Dement (Amst)2022; 14: e12374.
33.
WangCPaikMC. A weighting approach for GEE analysis with missing data. Commun Stat Theory Methods2011; 40: 2397–2411.
34.
AustinPCStuartEA. The performance of inverse probability of treatment weighting and full matching on the propensity score in the presence of model misspecification when estimating the effect of treatment on survival outcomes. Stat Methods Med Res2017; 26: 1654–1670.
