Causal Inference
Causal Diagrams for Disease Latency Bias Ramin Rezaeianzadeh* Mahyar Etminan Ramin Rezaeianzadeh Mohammad Mansournia
Objective: Disease latency is defined as the time from disease initiation to disease detection. Disease latency bias (DLB) can affect epidemiological studies that examine the causal effects of different exposures (eg, a drug) with a wide range of chronic diseases. Using casual diagrams we demonstrate four scenarios where disease latency can introduce bias into causal epidemiologic studies.
Methods: A number of epidemiologic studies have shown that benzodiazepines can increase the risk of dementia. Some of these results could have been affected by DLB since the prodromal signs of dementia could have preceded benzodiazepine use. We show 4 different causal directed acyclic graphs (cDAGs). We define variables: E (benzodiazepine use), Y (diagnosed dementia), Y* (early symptoms of dementia), U (unmeasured confounder) C are subjects censored from the study and M is a mediator between Y* and E.
Figure 1. Biasing path through an unmeasured confounder. U represents the unmeasured confounder ‘insomnia’. DLB might be introduced when U is a common cause of Y* (early signs of cognitive deficit secondary to dementia) and use of a benzodiazepine. A biasing path can be introduced through the path.
Y← *Y← U→ E
Figure 2. Biasing path through reverse causality bias. Early signs of cognitive deficit years prior to diagnosis of dementia can lead to use of benzodiazepines prior to dementia diagnosis. A biasing path is created through Y* which acts as an unmeasured confounder Y← Y*→E
Figure 3. Biasing path through Selection bias. C is a collider on the path E→ C←Y*→Y. Subjects who experience early cognitive adverse events of dementia (Y*) and an experience an adverse event of benzodiazepines are censored from the study and analysis is done only among those who stay in the study (represented with a boxed C=0).
Figure 4. Biasing path through a mediator. The effect of Y* is mediated on E through M (number of physician visits). Early signs of dementia prompts patients to have more physician visits which will make it more likely for them to be prescribed a benzodiazepine. Adjustment for M in these studies can lead to
Conclusion: Disease latency bias is am important bias that might effect epidemiologic studies that examine a latent outcome. Sensitivity analysis including probabilistic bias analysis can be undertaken to control for this bias.