The Opioid Crisis Had the Same Structural Topology as the Antidepressant Evidence Chain

The Question

We found structural circularity in the antidepressant evidence chain (M_MED1): the monoamine hypothesis was the top hub in every epistemic layer -- justification, endpoint selection, and citation support. Same node. Every layer. The founding assumption justified the research program, selected the endpoints, and was also the most-cited evidence for the research program.

The question is whether that pattern is specific to antidepressants, or whether it appears in other drug programs with known evidence quality problems. We already tested the healthy case (M_MED2: vaccines -- non-circular, different hubs in different layers). M_MED3 tests the circular case in a different domain: the opioid prescribing epidemic.

The opioid crisis is well-documented. A founding claim (chronic pain is undertreated), a misused citation (Porter & Jick 1980), short-term RCT data extrapolated to long-term chronic use, institutional campaigns (fifth vital sign), and prescribing guidelines that reinforced the original claim. We pre-registered four hypotheses and ran the structural analysis.

Pre-Registration

Pre-registered before any analysis was run.
• Hash: 3a14a005 (full: 3a14a005218c2384c7977769bad59f23c5516390046d59289c8461ffa257c793)
• Timestamp: 2026-06-02T18:10:04 UTC
• Public record: https://github.com/vladi160/preregistrations/commit/caa26e2

The Dataset

Eight nodes representing the opioid prescribing evidence chain:
• pain_undertreated_claim -- founding claim: chronic pain is systematically undertreated
• mu_opioid_receptor_mechanism -- opioids bind mu-receptors, reducing pain transmission
• vas_pain_scale -- Visual Analog Scale (patient self-report, 0-10)
• porter_jick_letter -- 1980 NEJM 5-sentence letter; 0.03% addiction in hospitalized patients; misused for chronic outpatient use
• short_term_rct_efficacy -- RCTs showing VAS reduction vs placebo (mostly 6-12 weeks)
• fda_chronic_pain_approval -- FDA approval for chronic non-cancer pain (extrapolated from short-term data)
• fifth_vital_sign_jcaho -- JCAHO campaign: pain as the fifth vital sign; mandated pain assessment
• prescribing_guidelines_aps -- APS/JCAHO guidelines mandating aggressive pain treatment

Three relational layers: justifies (d1), selects_endpoints (d2), cites_as_support (d3).

What IRDME Found

h1 -- FPL gradient: PARTIAL r(justifies, selects_endpoints) = 0.1443, Spearman = 0.376, p = 0.862. Direction positive (correct) but below the r >= 0.20 threshold and not statistically significant. The FPL gradient collapses in this circular chain.

h2 -- pain_undertreated_claim is the justification hub: CONFIRMED Rank #1 in justifies by both degree and betweenness centrality. The most structurally central node in the full graph. It justifies the mu-opioid mechanism, the fifth vital sign campaign, and the prescribing guidelines directly. It also receives a circular edge back from the guidelines.

h3 -- pain_undertreated_claim is ALSO the citation hub: CONFIRMED Rank #1 in cites_as_support. The same founding claim that dominates justification also dominates citation. porter_jick_letter cites it, short-term RCT results cite it, prescribing guidelines cite it in a circular loop.

This is the structural circularity signature: the founding claim stays at the top across all epistemic layers. It cannot be escaped -- it justifies things and is also cited as evidence for things. The evidence chain does not reach an independent empirical anchor outside the founding assumption.

h4 -- porter_jick_letter citation anomaly: CONFIRMED Rank #3 in cites_as_support. Peripheral in justifies (only one outgoing justification edge: to mu_opioid_receptor_mechanism). IRDME detected this anomaly without reading any paper.

Key Finding

Structural circularity confirmed, replicated across drug classes.

The opioid evidence chain and the antidepressant evidence chain show structurally isomorphic patterns:
• Antidepressants: monoamine_hypothesis rank #1 in justifies AND cites_as_support
• Opioids: pain_undertreated_claim rank #1 in justifies AND cites_as_support
• Vaccines (healthy control): adaptive_immunity_mechanism rank #1 in justifies; clinical_vaccination_guidelines rank #1 in cites_as_support (DIFFERENT nodes)

The Porter & Jick anomaly. The 1980 NEJM letter (Porter & Jick) is a famous case of citation misuse: a 5-sentence letter about 11,882 hospitalized patients was cited thousands of times as evidence that chronic outpatient opioid use carries low addiction risk. IRDME flags this without reading a word. The structural signature: rank #3 in cites_as_support (heavily cited as evidence), but peripheral in justifies (provides almost no theoretical justification). A node that is cited more than it justifies is a diverger between the epistemic layers -- the structural fingerprint of a misused citation.

FPL gradient as a circularity sensor. Comparing the three experiments:
• M_MED2 vaccines (healthy): r = 0.75, p = 0.014 -- CONFIRMED
• M_MED1 antidepressants (circular): r = 0.41, p = 0.44 -- not significant
• M_MED3 opioids (circular): r = 0.14, p = 0.86 -- not significant, gradient collapse

The pattern: circular evidence chains show weak or non-significant FPL gradients. Healthy chains show strong gradients. The structural interpretation: when the same node dominates all layers (circularity), the layer structure becomes homogeneous -- there is no distinct epistemic separation between justification and citation, so the gradient collapses.

Interpretation

Both the antidepressant and opioid crises share a structural signature that predates public awareness of their problems. A structural analysis tool in 1990 or 2000 would have seen the same topology we see now -- the same node dominating the same layers. The topology does not need hindsight.

What this does not claim: the structural pattern does not explain WHY the evidence is circular (industry influence, publication bias, endpoint selection, regulatory extrapolation). It identifies that the structure is circular and flags the anomalous nodes. The mechanistic explanation requires domain-specific knowledge that IRDME does not supply.

What it does claim: the topology of a pharmaceutical evidence chain, encoded as a multilayer graph with three epistemic layers, carries a structural signal that distinguishes healthy chains from circular ones -- detectable from graph topology alone without reading any paper content.

Limitations and Next Steps

n=8 models the key structural claims; not exhaustive. Porter & Jick being flagged is robust to this choice (it would be flagged in any reasonable graph that includes it as a cited source).

The three-experiment series is provisional. Formal validation requires testing across 10+ drug classes with known replication outcomes and computing a circularity score.

The next test: does a circular score (rank of founding_claim in cites_as_support) correlate with eventual replication failure rates when applied prospectively?

M_MED4 is the next candidate: schizophrenia (dopamine/glutamate/NMDA multilayer) -- predict higher circularity than vaccines, potentially lower than opioids.

Pre-Registration Record

Experiment: M_MED3_v1

Dataset: opioid_evidence_chain.json, n=8, 3 layers, 22 relations

Hash: 3a14a005218c2384c7977769bad59f23c5516390046d59289c8461ffa257c793

Timestamp: 2026-06-02T18:10:04.447630+00:00

Commit: https://github.com/vladi160/preregistrations/commit/caa26e2

Results: h1 PARTIAL r=0.1443 p=0.862 (direction correct, below threshold), h2 CONFIRMED (pain_undertreated_claim rank #1 justifies), h3 CONFIRMED (pain_undertreated_claim rank #1 cites_as_support), h4 CONFIRMED (porter_jick_letter rank #3 cites_as_support)

Verified: hypotheses unchanged, hash intact