Untangling Causal Directionality Between Biomarkers and Disease

51 concepts51 papers1 proposal

When a clinical study finds that a specific molecule in the blood correlates with a disease, a critical question arises: Is this molecule a driver or a passenger? In epidemiology and molecular medicine, this is known as the problem of causal-correlational ambiguity or reverse causality.

If a biomarker is a direct cause of a disease, intervening to change its levels can prevent or treat the condition. If it is merely a consequence—a reactive symptom or a downstream byproduct—altering it is clinically useless. Distinguishing between these two scenarios is one of the most significant challenges in modern biomedical research.

The Intuition of Reverse Causality

To understand reverse causality, imagine observing that fire trucks are highly correlated with house fires. A naive observer might conclude that fire trucks cause fires. In medicine, we frequently face the same trap.

For instance, Lijun Zhang et al. noted that lower circulating cholesterol levels are associated with an increased risk of Parkinson's disease. Does low cholesterol drive neurodegeneration, or does the early, undiagnosed stage of the disease alter systemic metabolism and lifestyle, thereby lowering cholesterol?

Similarly, Luo Yuxin et al. highlighted the long-standing debate over whether elevated serum uric acid levels actively drive chronic kidney disease or simply serve as a biomarker of existing kidney damage that has reduced the organ's clearance capacity.

When a condition itself triggers physiological, metabolic, or behavioral changes, traditional observational studies cannot determine which came first.

Observational Correlation:
[ Biomarker ] <--- Ambiguous Relationship ---> [ Disease State ]

Causal Scenario:
[ Biomarker ] ------------------------------> [ Disease State ]

Reverse Causality Scenario:
[ Biomarker ] <------------------------------ [ Disease State ]

Why Observational Data Fails Us

Why can't traditional studies tell these scenarios apart? Cross-sectional and observational studies capture a snapshot in time where both the disease and the biomarker coexist.

Whether researchers look at physiological signals like heart rate variability (Masayuki Yotsukura et al.), circulating antibodies (Suvi Μ. Virtanen et al.), or complex epigenetic modifications like DNA methylation (Lang Wu et al.; Yanfa Sun et al.), the biomarker and the pathology are deeply entangled.

A molecular signature might look like a perfect early-warning signal—such as the microRNA signatures studied by Bilal Khan et al. or the plasma cell-free RNA signatures identified by Li Wu et al.—but still be a passive downstream indicator of existing cellular damage rather than an active participant in the disease mechanism. As Shuai Liu et al. point out, the disease process itself frequently alters protein expression, confounding observational protein-disease associations.

Mendelian Randomization as a Causal Anchor

To break this deadlock, researchers leverage a method called Mendelian Randomization (MR). MR acts as a natural randomized controlled trial by using genetics to anchor the direction of causality.

At conception, nature randomly assigns genetic variants (alleles) that influence our baseline biological traits, such as the level of a specific blood metabolite or plasma protein. Because our inherited DNA sequence is fixed at birth and cannot be altered by a disease that develops decades later, these genetic variants serve as unconfounded "instruments" to establish the direction of cause and effect.

Nature's Randomized Trial:
[ Genetic Variant (Fixed at Birth) ] ---> [ Baseline Biomarker Level ] ---> [ Disease Risk ]

If a genetic variant known to increase a biomarker also increases disease risk, the arrow of causality must flow from the biomarker to the disease.

Using this logic, Sarah Filippi applied MR to demonstrate a causal link between high diastolic blood pressure and accelerated brain aging. Similarly, Qinghao Zhao et al. used MR to isolate fibroblast growth factor 23 (FGF23) as a primary causal driver of aortic stenosis, distinguishing it from other mineral markers that are merely correlated. Jia-Lin Ren also leveraged genomic data to identify specific blood metabolites that actively drive or protect against multiple myeloma, bypassing external lifestyle confounders.

Bidirectional MR: Mapping the Feedback Loops

Sometimes, the relationship isn't a one-way street; a biomarker might cause a disease, and the disease might simultaneously alter the biomarker. To map these complex dynamics, researchers employ bidirectional Mendelian Randomization, testing the causal arrow in both directions using separate sets of genetic instruments.

Bidirectional Analysis:
Forward Path:  [ Genetic Instruments for Biomarker ] -------> [ Biomarker ] -------> [ Disease ]
Reverse Path:  [ Genetic Instruments for Disease ] ---------> [ Disease ] ---------> [ Biomarker ]

Consider the study by Kun‐Ju Lin et al., which investigated the plasma protein SPAG11B and rheumatoid arthritis (RA). Their bidirectional analysis revealed a striking dual relationship: genetically higher levels of SPAG11B protect against RA, but the presence of RA conversely acts to lower SPAG11B levels. Without a bidirectional approach, this feedback loop would appear as a confusing, uninterpretable correlation.

Bidirectional MR allows researchers to systematically untangle these relationships across various domains:

  • Osteoporosis: Yimin Liu et al. evaluated 35 blood and urine metabolites against postmenopausal osteoporosis to separate true upstream drivers from metabolic consequences of bone loss.
  • Tissue Degeneration: Yao Wang et al. proved that specific systemic inflammatory proteins are upstream drivers of tendon degeneration rather than just a secondary response to mechanical damage.
  • Metabolic Complications: Bo Li et al. applied bidirectional MR to distinguish which circulating metabolites are true causal drivers of diabetic retinopathy versus those that are merely biomarkers of its progression.

Moving from Observation to Intervention

Distinguishing a cause from a consequence is the difference between a failed clinical trial and a successful drug. If a protein is merely a reactive biomarker—such as the oxidative stress markers identified by Shuli Li or the iron signature studied by Rong Hu et al.—pharmaceuticals designed to alter its levels will fail to improve clinical outcomes. As noted by the authors of a large-scale proteomic study on Alzheimer's disease risk, isolating causal relationships prevents researchers from targeting proteins that are merely symptoms of a disease.

Conversely, when MR or functional validation confirms a causal role, it pinpoints high-confidence therapeutic targets. For instance, Lihua Wang et al. used MR to show that higher soluble TREM2 levels are causally protective against Alzheimer’s disease, transforming it from a vague neuroinflammatory marker into a concrete target for therapeutic upregulation.

When Haole Liu et al. used CRISPR-modified mice to eliminate C-reactive protein (CRP), they proved that CRP is an active participant in tissue destruction rather than a passive bystander. By moving beyond simple correlation, these methods allow medicine to transition from passive diagnostic observation to targeted therapeutic intervention.

Key Foundational Work

Heart rate variability within 24 hour after the onset of acute myocardial infarction. (1996) — contributed 1 concept elements to this direction.
Diet, cow's milk protein antibodies and the risk of IDDM in Finnish children (1994) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_reverse_causality_3744d154,
  title = {reverse causality},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/reverse-causality/rdr_3744d15485ae49888e5913a69c76aecc},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_reverse_causality_3744d154_v1,
  title = {reverse causality},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/reverse-causality/rdr_3744d15485ae49888e5913a69c76aecc/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake