2015 Gesundheitswesen (Bundesverband der Arzte des Offentlichen Gesundheitsdienstes (Germany))
* [Time-dependent confounding in the estimation of treatment effects in randomised trials with multimodal therapies--an illustration of the problem of time-dependent confounding by causal graphs].
- Biased effect estimates induced by unconsidered confounding variables are a known problem in observational studies. Selection bias, resulting from non-random sampling of study participants, is widely recognised as a problem in case-control and cross-sectional studies. In contrast, possible bias in randomised controlled trials (RCTs) is mostly ignored. This paper illustrates, by applying directed acyclic graphs (DAGs), possible bias in the effect estimates of first-line therapy, caused by subsequent changes in therapy (time-dependent confounding). Possible selection bias, induced by not only random loss to follow-up, will be explained as well using DAGs. Underlying assumptions of standard methods usually used to analyse RCTs (like intention-to-treat, per-protocol) are shown and it is explained why effect estimates may be biased in RCTs, if only these conventional methods are used. Adequate statistical methods (causal inference models as marginal structural models and structural nested models) exist. Higher documentary efforts, however, are necessary, because any changes in medication, loss to follow-up as well as reasons for such changes need to be documented in detail as required by these advanced statistical methods. Nevertheless, causal inference models should become standard along side the currently applied standard methods, especially in studies with high non-compliance due to changes in therapy and substantial loss to follow-up. Possible bias cannot be excluded if similar results are obtained from both methods. However, study results should be interpreted with caution if they differ between both approaches.
=>趣旨, 結果, 影響, 効力, 活動, 印象, 効果, 成し遂げる
Overview of noun effect
The noun effect has 6 senses (first 5 from tagged texts)
1. (101) consequence, effect, outcome, result, event, issue, upshot -- (a phenomenon that follows
and is caused by some previous phenomenon; "the magnetic effect was greater when the rod was
lengthwise"; "his decision had depressing consequences for business"; "he acted very wise after the
event")
2. (11) impression, effect -- (an outward appearance; "he made a good impression"; "I wanted to
create an impression of success"; "she retained that bold effect in her reproductions of the
original painting")
3. (9) effect -- (an impression (especially one that is artificial or contrived); "he just did it
for effect")
4. (2) effect, essence, burden, core, gist -- (the central meaning or theme of a speech or literary
work)
5. (1) effect, force -- ((of a law) having legal validity; "the law is still in effect")
6. effect -- (a symptom caused by an illness or a drug; "the effects of sleep loss"; "the effect of
the anesthetic")
Overview of verb effect
The verb effect has 2 senses (first 2 from tagged texts)
1. (17) effect, effectuate, set up -- (produce; "The scientists set up a shock wave")
2. (3) effect -- (act so as to bring into existence; "effect a change")
--- WordNet end ---