A scoping review of survival analysis methods for doubly interval censored data

Doubly interval-censored data (DICD) often emerge in longitudinal studies, where the survival time, T = W − V, denotes the duration between two connected events: the initial event time (V) and the subsequent event time (W), both of which are subject to interval censoring (IC). Such data are prevalent in medical and epidemiological research, especially when event occurrences are observed only within specific intervals because of the nature of study designs or periodic assessments. Analysing DICD poses unique challenges: the lack of exact event times introduces complexities in modelling, estimation, and inference, which leads to biased regression parameter estimates and higher standard errors, if not handled appropriately. Existing methods often struggle with computational demands and theoretical constraints, particularly in scenarios involving small sample sizes or irregular censoring intervals. This study provides a comprehensive overview of the methods developed to analyse DICD, examining their strengths, limitations, and practical applications. We systematically reviewed the literature to achieve this goal, focusing on studies retrieved from PUBMED, Scopus, Embase, and CINAHL from database inception through June 2024. Among the 462 screened studies, 52 met the inclusion criteria, with applications of DICD most commonly arising in medical research. We distilled 8 key topic areas based on the synthesis of available literature and underscore the importance of understanding and appropriately analysing DICD to ensure accurate inferences and informed decision-making across various scientific disciplines. This review highlights how authors justify the use of DICD, the methods employed, and the limitations addressed, offering valuable insights into this specialized area of survival analysis.