Immortal Time Bias refers to a bias in observational studies with time-to-event endpoints, where a treatment group is credited with a period of time during which no event can occur. A typical example is the time between cohort entry and the first prescription: to be considered treated later, an individual must remain event-free until this prescription. The artificially created survival advantage can then appear to be a treatment effect. <\/p>\n
The error occurs when exposure is defined based on information from the future, but follow-up begins earlier. For example, if a comparison is made from the date of diagnosis to determine who ever received therapy, the days before the start of therapy are retrospectively assigned to the treated group. A death in this interval prevents the later prescription and therefore automatically assigns the person to the untreated group. <\/p>\n
The bias does not only affect deaths. Hospitalization, progression, or a composite endpoint also cannot occur within the period in question if their occurrence makes later exposure classification impossible. The longer the period until the start of treatment and the more frequent the event, the more the naive classification can shift the effect towards a treatment advantage. <\/p>\n
Time-dependent modeling initially treats individuals as unexposed and only allows their status to change at the actual start of treatment. This approach is suitable when the scientific question concerns the initiation of therapy during ongoing observation, and the timing and status changes are reliably available. A mere \u201cever treated\u201d variable in the initial model does not correct the error. <\/p>\n
In a landmark analysis, only individuals who are event-free and observed up to a pre-defined landmark time point are included; the exposure status is determined at this point. Prescription Time-Distribution Matching, on the other hand, uses the observed distribution of prescription times to assign a corresponding index time point to comparison individuals. Both approaches answer a more narrowly defined question and must disclose the resulting altered target population. <\/p>\n
Defining a minimum duration of therapy can also create immortal time. If only individuals who achieve three prescriptions are considered exposed, they must necessarily survive the interval until the third dispensing without an endpoint. In this case, the comparison must represent exposure as a time-varying, dynamic strategy or set the starting point for both groups to a clinically justified landmark. <\/p>\n
Immortal Time Bias is neither a variant of selection bias nor a recall bias. It is based on an incorrect temporal alignment of exposure definition and the start of observation. The bias can therefore occur even if diagnoses and prescriptions have been fully recorded and the treated and untreated individuals appear similar at cohort entry. <\/p>\n
The existing glossary term \u201cbias\u201d describes distortion generally; Immortal Time Bias, however, specifically names this erroneous assignment of event-free time. It is distinct from confounding: adjustment for age, severity, or co-medication does not correct for days that were incorrectly credited to the wrong exposure group. First, the risk time must be correctly constructed, and only then can confounder methods be applied. <\/p>\n
For retrospective registry and chart studies, the protocol and SAP should therefore separately record the index date per comparison group, the rule for starting a treatment, and any allowed status changes. Analyses after a later intervention, such as consolidation therapy or an intervention after response, are particularly critical. A data table must allow verification that events before the first prescription count as unexposed risk time and not as survival time for the later treated group. <\/p>\n
An effective quality check compares each individual’s diagnosis, index, prescription, event, and censoring dates in their actual sequence. Conspicuous negative intervals or events before the declared start of exposure are not minor data errors but can indicate that the chosen analysis timeline does not represent the underlying therapeutic process. <\/p>\n
Full-service CROs like Mediconomics support in avoiding Immortal Time Bias by creating timelines, specifying time-dependent exposures in the analysis plan, deriving landmark rules, and programming checks of person-period data. For study reports, the chosen index data, exclusions at the landmark, and the results of alternative timely analyses can be transparently described. <\/p>\n
When is there a concrete suspicion of Immortal Time Bias?<\/strong><\/p>\n Suspicion arises when an individual can only receive treatment after cohort entry but is counted as treated from cohort entry. The check then asks whether an event before the first treatment would preclude later assignment to the treated group. <\/p>\n Is it sufficient to start follow-up at the first prescription?<\/strong><\/p>\n This may be appropriate for a comparison of new users with suitable non-users defined at the same time point. However, without an equivalent index rule for the comparison group, this choice merely shifts the problem and can create new selection differences. <\/p>\n Which correction is always the best?<\/strong><\/p>\n No method is universally superior. Time-dependent models, landmark analyses, and Prescription Time-Distribution Matching presuppose different target populations and comparison questions. The choice follows the study objective, the availability of precise time points, and the clinical significance of a fixed landmark time point. <\/p>\n Immortal Time Bias refers to a bias in observational studies with time-to-event endpoints, where a treatment group is credited with a period of time during which no event can occur. A typical example is the time between cohort entry and the first prescription: to be considered treated later, an individual must remain event-free until this […]<\/p>\n","protected":false},"author":10,"featured_media":0,"parent":0,"template":"","meta":{"_acf_changed":false,"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"glossary-cat":[22],"class_list":["post-7808","glossary","type-glossary","status-publish","hentry","glossary-cat-biostatistik-methodik"],"acf":[],"related_terms":"","external_url":"","internal_reference_id":"","_links":{"self":[{"href":"https:\/\/mediconomics.com\/en\/wp-json\/wp\/v2\/glossary\/7808","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mediconomics.com\/en\/wp-json\/wp\/v2\/glossary"}],"about":[{"href":"https:\/\/mediconomics.com\/en\/wp-json\/wp\/v2\/types\/glossary"}],"author":[{"embeddable":true,"href":"https:\/\/mediconomics.com\/en\/wp-json\/wp\/v2\/users\/10"}],"version-history":[{"count":0,"href":"https:\/\/mediconomics.com\/en\/wp-json\/wp\/v2\/glossary\/7808\/revisions"}],"wp:attachment":[{"href":"https:\/\/mediconomics.com\/en\/wp-json\/wp\/v2\/media?parent=7808"}],"wp:term":[{"taxonomy":"glossary-cat","embeddable":true,"href":"https:\/\/mediconomics.com\/en\/wp-json\/wp\/v2\/glossary-cat?post=7808"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}Regulatory References<\/h2>\n
\n