Farb et al. (2022) — Static and dynamic brain biomarkers of depression relapse
The RCT sequel to Farb et al. (2011), and the paper that turns the 2011 correlation into a two-factor model with an intervention attached. Where 2011 followed 16 medicated patients for 18 months and found that the mode of cortical reactivity to a sad film forecasts relapse, this scans 85 remitted patients before and after eight weeks of prophylactic psychotherapy, follows them for two years, and asks the question 2011 could not: which part of the vulnerability signature is a fixed trait, and which part does treatment actually move? Ingested from Week 11 The Absence of Emotion — the second of three Farb papers in that folder (after 2011, before the 2010 sadness/MBSR paper).
Design
Remitted, recurrent-MDD outpatients from a larger RCT (n=166 randomized; NCT01178424) were assigned to MBCT (Segal, Williams & Teasdale) or Well-Being Cognitive Therapy (WB-CT; Fava) — two evidence-based relapse-prophylaxis programmes with divergent procedures. 85 completed both a pre- and a post-intervention fMRI scan of the same validated sad-vs-neutral film dysphoric mood induction used in 2010/2011, then entered a 24-month clinical follow-up (primary outcome: time to DSM-IV MDE by SCID + HRSD>=16). Eighteen relapsed. Because each patient was scanned twice, reactivity common to both scans estimates a static (trait-like) marker, and reactivity change across scans estimates a dynamic (treatment-responsive) marker — the design’s central move. All fMRI models and Cox regressions controlled for past episodes, residual symptoms, and antidepressant status.
The two factors
| factor | region | direction in relapsers | treatment moves it? | hazard ratio |
|---|---|---|---|---|
| static (trait vulnerability) | right somatosensory cortex (+ bilateral S1, SMA, fusiform) | greater sadness-evoked deactivation | no — identical pre/post | HR 0.039 [0.011, 0.14], p<.001 |
| dynamic (treatment target) | left DLPFC / lateral PFC | failure to reduce reactivity over treatment | yes — drops in non-relapsers only | HR 3.73 [1.33, 10.46], p=.013 |
Both survived in a combined Cox model controlling for the three clinical covariates, together accounting for 89.7% of relapse variance (C = 0.86). They are the same two poles as 2011 — sensory deactivation (harmful when it goes down) and prefrontal cognitive reactivity (harmful when it stays up) — now separated by their response to intervention.
The static factor is the one 2011 found, replicated and extended
Greater sadness-evoked sensory deactivation marks relapse risk, is present equally before and after therapy, and prophylactic treatment does not touch it. This both replicates the 2011 sensory-deactivation direction and delivers the paper’s most sobering clinical result: the trait vulnerability the field would most want to fix is the one neither MBCT nor WB-CT moved. The covariate analyses seat it deeper — more past episodes and worse residual symptoms both predict more of this same somatosensory + insula deactivation (replicating Farb et al. 2010) — so sensory deactivation indexes past, present, and future depression at once.
The dynamic factor is what therapy actually does
Left-DLPFC reactivity fell over the eight weeks in patients who stayed well and did not in patients who relapsed, most visibly in those with high baseline prefrontal reactivity. This is the treatment-modifiable half of the model: prophylaxis works (to the extent it works) by turning down maladaptive prefrontal cognitive reactivity, not by restoring sensory integration. It is the cognitive-reactivity construct with a longitudinal, treatment-responsive brain correlate.
Trans-therapeutic: mindfulness is not privileged here
Aim 3 sought MBCT-vs-WB-CT mechanistic differences and found none — no Group x Time, no Group x Time x Relapse, no relapse-rate difference. Despite emphasizing divergent strategies (mindful decentering vs. well-being reappraisal), both therapies converge on the same prophylactic marker: reduced DLPFC reactivity. This matches the parent trial’s null process dissociation (Segal et al. 2019) and matters for this wiki’s applied pages: the prefrontal mechanism is common to the therapies, so the reductions cannot be attributed to interoceptive/mindfulness training specifically. See mindfulness-interoceptive-training, does-mindfulness-enhance-interoceptive-accuracy.
The PPI: elaboration suppresses sensation
Seeding the right somatosensory static marker, dysphoric mood flipped its connectivity with the right lateral PFC from positive (neutral mood) to negative (sad mood) — an inhibitory, contralateral mirror of the left-LPFC dynamic marker — and the size of that positive→negative flip tracked relapse. Read structurally, this is the mechanism the whole two-factor model implies: under dysphoria, prefrontal cognitive elaboration actively inhibits sensory integration. Depression vulnerability is not two independent lesions but a single over-reliance on cognitive elaboration at the expense of sensory representation — the active-inference (elaborate/regulate) vs. perceptual-inference (stay perceptual) trade-off Farb’s interoception theory is built on, here with a hard clinical outcome and a directed connection.
The honest limit for an interoception wiki (updated from 2011)
2011’s protective sensory pole was purely visual (calcarine cortex) — exteroceptive, and the wiki flagged the interoceptive reading as analogy. 2022 moves the sensory pole closer to interoception but does not land on it:
- The static marker is dominated by primary somatosensory cortex, SMA, and fusiform — somatomotor and visual, i.e. still largely exteroceptive. Right S1, the strongest single predictor, is exteroceptive touch, not interoception.
- The insula — the interoceptive cortex proper (insular-cortex, Craig 2002) — appears in the covariate maps (anterior + posterior insula deactivation scaling with past episodes and residual symptoms) and in the paper’s framing of “sensory deactivation” as including “awareness of the body’s internal state,” but it is not the strongest relapse predictor here.
- The paper’s own frame is broad “sensory integration” deactivation across somatosensory, motor, insular, and visual cortices — of which interoception (insula) is one component, not the headline.
So the interoceptive claim is still a generalization — but a better-supported one than 2011, because the deactivating territory now includes somatosensory and insular cortex rather than vision alone. What the paper establishes cleanly is the shape the interoception theory borrows: dysphoric elaboration suppresses bodily/sensory representation, and that suppression forecasts illness return. See two-factor-model-of-relapse-vulnerability, is-more-interoceptive-awareness-better.
Provenance
Norman Farb is this wiki’s author (see norman-farb); recorded as provenance, not weighted as authority — the convention used for farb-2011-relapse-prediction and payne-2015-somatic-experiencing. In scope because only raw/papers/Farb/ is excluded from auto-ingest; this sits in a course-week folder. Senior author Segal co-developed MBCT and declares related book/workshop revenue (Farb, Anderson, Desormeau declare none). Not a duplicate of farb-2011-relapse-prediction (different sample, RCT design, journal, and the static/dynamic decomposition that is new here) or farb-2015-interoception-contemplative-health (theory paper). Second author Philip Desormeau is given no researcher page — co-author, not independently load-bearing here.