Landmarking for Left‐Truncated Competing Risk Data

ABSTRACT Landmarking is an alternative to complex multistate models when the aim is to calculate dynamic predictions. We develop the concept of landmarking for the case of left truncation and competing risks from the application background of drug safety assessment in pregnancy. The method is illust...

Full description

Saved in:
Bibliographic Details
Published in:Biometrical journal 2024-12, Vol.66 (8), p.e202400083-n/a
Main Authors: Unseld, Theresa, Bluhmki, Tobias, Beyersmann, Jan, Beck, Evelin, Padberg, Stephanie, Stegherr, Regina
Format: Article
Language:English
Subjects:
Abortion, Spontaneous - chemically induced
Abortion, Spontaneous - epidemiology
Biometry - methods
Cohort Studies
competing risks
Estimates
Exposure
Female
Fluoroquinolones - adverse effects
Humans
landmarking
left truncation
Models, Statistical
Pharmacovigilance
Pregnancy
pregnancy data
Regression analysis
Risk
Risk Assessment
time‐dependent covariates
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:ABSTRACT Landmarking is an alternative to complex multistate models when the aim is to calculate dynamic predictions. We develop the concept of landmarking for the case of left truncation and competing risks from the application background of drug safety assessment in pregnancy. The method is illustrated with a cohort study of the German Embryotox Pharmacovigilance Institute in Berlin to assess if the risk or the cumulative incidence of adverse pregnancy outcomes, like spontaneous abortions (SABs), is increased in fluoroquinolone‐exposed women. Furthermore, we conduct an extensive simulation study to compare the dynamic predictions and coefficient estimates obtained by landmarking to those from nonparametric multistate models and classical time‐dependent covariate Cox regression. The results from the simulation study indicate that attenuation of the effects is present in the landmark estimates, also in the complex setting considered here, but the estimates are still close to those from the multistate models. Regarding the Berlin fluoroquinolone data, the fluoroquinolone exposure of a pregnant woman in the first trimester seems to increase her cumulative incidence of elective termination of pregnancy over women never exposed before, but there is no evidence of a significantly increased risk or cumulative incidence in exposed women for SABs. This supports previous results on the same data, which were driven from an analysis without landmarking methods.
ISSN:0323-3847
1521-4036
1521-4036
DOI:10.1002/bimj.202400083