Disentangling Bidirectional Causality in Chronic Diseases

47 concepts43 papers1 proposal

When two medical conditions frequently occur together, clinicians face a classic "chicken-or-egg" dilemma. Does condition A trigger condition B, does condition B cause condition A, or are both driven by a hidden third factor?

In traditional observational epidemiology, answering this question is notoriously difficult. If a study measures health markers at a single point in time, it cannot establish a temporal sequence. Even longitudinal studies can be misled by reverse causality or shared environmental confounders like socioeconomic status, diet, or aging.

To move beyond mere correlation, a growing body of research leverages advanced causal inference techniques—such as Mendelian randomization (MR) and longitudinal path analysis—to map the precise directional arrows driving complex comorbidities.

The Failure of Observational Symmetry

To understand why traditional approaches struggle, consider the relationship between muscle weakness and joint pain. In clinical settings, patients with "runner’s knee" (Patellofemoral Pain Syndrome, or PFPS) often exhibit profound hip weakness. For years, sports medicine assumed that weak hips caused poor running mechanics, leading to knee injury.

However, by studying athletes at the very inception of the condition, Christopher T. Plastaras et al. demonstrated in Is Hip Abduction Strength Asymmetry Present in Female Runners in the Early Stages of Patellofemoral Pain Syndrome? that early-stage patients showed no significant hip strength asymmetry. This strongly implies that the weakness observed in chronic cases is a consequence of pain-induced disuse rather than its initial trigger.

When data is collected cross-sectionally, this directional ambiguity leaves clinicians blind. If a physical symptom and a psychological symptom co-occur, treating the wrong one first can lead to ineffective interventions and wasted resources.

Observational Correlation:
[ Condition A ] <======== (Statistical Association) ========> [ Condition B ]
                                 ^
                                 | (Confounded by environment/genetics)
                          [ Hidden Factor X ]

True Causal Models:

1. Unidirectional:  [ Condition A ] -------------------------> [ Condition B ]
2. Reverse:         [ Condition A ] <------------------------- [ Condition B ]
3. Bidirectional:   [ Condition A ] <------------------------> [ Condition B ]

Breaking the Symmetry with Mendelian Randomization

To determine the true direction of the causal arrow, researchers increasingly rely on Mendelian randomization (MR). MR acts as nature’s randomized controlled trial by exploiting the way genetic variants are randomly distributed at conception.

The logic of MR relies on three strict assumptions:

  1. The genetic variant is strongly associated with the exposure (Condition A).
  2. The variant is not confounded by external lifestyle or environmental factors.
  3. The variant affects the outcome (Condition B) only through its effect on the exposure, not through an alternative biological pathway (a violation known as horizontal pleiotropy).
Mendelian Randomization Architecture:
[ Random Allele Allocation ] 
            |
            v
     [ Exposure (A) ] ---------------> [ Outcome (B) ]
            |                                 ^
            +-------- (No Direct Path) -------+

Because an individual's DNA sequence is fixed at birth, it cannot be altered by the subsequent onset of disease. Therefore, if genetic variants that specifically predispose a person to Condition A also reliably predict a higher risk of developing Condition B, the causal arrow must point from A to B. Reverse causality is mathematically ruled out.

Using this framework, Rui Lai et al. investigated the co-occurrence of autoimmune and endocrine disorders in Causal relationship between rheumatoid arthritis and hypothyroidism or hyperthyroidism: a bidirectional two-sample univariable and multivariable Mendelian randomization study. While observational data simply linked the two, the authors' bidirectional MR analysis proved a nuanced, asymmetrical relationship: a genetic liability for rheumatoid arthritis causally increases the risk of hypothyroidism, and vice versa. However, while rheumatoid arthritis also causally drives hyperthyroidism, the reverse path does not hold.

True Feedback Loops vs. Unidirectional Cascades

When these causal tools are applied globally, they reveal that comorbidities generally fall into two distinct structural architectures: true bidirectional feedback loops and unidirectional cascades.

1. True Bidirectional Feedback Loops

In these systems, two conditions act as a mutually reinforcing downward spiral. A clear example is the brain-gut axis. Patients with neuropsychiatric conditions frequently suffer from gastrointestinal distress.

In Neuropsychiatric Disorders and Constipation: Unraveling Causal Links Through Genetic Analysis, Guojie Zhao et al. demonstrated a genuine two-way genetic street. Neuropsychiatric disorders causally alter gut motility, while the physiological stress of chronic constipation simultaneously feeds back to worsen psychiatric outcomes. A similar bidirectional relationship was mapped by Xuefei Wang et al. in The causal interplay between depression and alcohol use from adolescence to young adulthood: a Mendelian randomization study, showing how early-life depression drives alcohol misuse, which subsequently alters neural chemistry to deepen depressive states.

2. Unidirectional Cascades

Conversely, some relationships that appear bidirectional in a clinic are actually strictly one-way paths. Consider obstructive sleep apnea (OSA) and atrial fibrillation (AF), a common heart rhythm disorder.

By applying a bidirectional MR approach, Xingang Sun et al. showed in Obstructive sleep apnea and atrial fibrillation: insights from a bidirectional Mendelian randomization study that genetic predisposition to OSA directly increases AF risk, likely due to the mechanical and hypoxic stress placed on the heart during nighttime airway collapses. Crucially, the reverse path was entirely non-significant: having a genetic risk for heart arrhythmias does not cause sleep apnea.

Causal Discovery Without Genetics

While MR is highly effective, it requires robust genome-wide association data, which is not always available for every clinical symptom. In these scenarios, researchers turn to computational causal discovery and path analysis applied to longitudinal datasets.

Instead of relying on genetics, these methods analyze the "arrow of time" across repeated measurements. If changes in symptom A at Time 1 consistently predict changes in symptom B at Time 2, while changes in symptom B at Time 1 fail to predict symptom A at Time 2, a directional inference can be made.

Using longitudinal causal discovery across five distinct patient populations, Benjamin Pierce et al. published Causal discovery replicates symptomatic and functional interrelations of posttraumatic stress across five patient populations. They proved that what clinicians often treat as independent, co-occurring disorders—such as PTSD and major depression—is actually a directional cascade. PTSD symptoms consistently act as the primary engine, directly driving downstream depressive symptoms and functional impairment.

Why the Distinction Matters for Treatment

Unmasking the true direction of causality fundamentally changes how medicine is practiced.

If two conditions are merely correlated due to a shared confounder, treating condition A will do nothing to prevent condition B. If the relationship is a unidirectional cascade, treating the true root cause (e.g., managing sleep apnea) can successfully prevent the downstream complication (e.g., atrial fibrillation). Finally, if the system is a true bidirectional loop, clinicians know they cannot rely on a single silver-bullet treatment; instead, they must deploy simultaneous, dual-target interventions to break the cycle.

By moving away from simple correlation and mapping the directional architecture of disease, these methodologies allow public health and clinical medicine to transition from reactive symptom management to precise, causal prevention.

History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_bidirectional_causality_bidirectional_he_4c3396af,
  title = {bidirectional causality (bidirectional health causality)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/bidirectional-causality-bidirectional-health-causality/rdr_4c3396af73ca4ffaa1972b636fb23e9b},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_bidirectional_causality_bidirectional_he_4c3396af_v1,
  title = {bidirectional causality (bidirectional health causality)},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/bidirectional-causality-bidirectional-health-causality/rdr_4c3396af73ca4ffaa1972b636fb23e9b/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake