•Expectancies modulate both reported pain and responses in some brain regions, yet the key brain circuitry that mediates expectancy effects on pain experience has not been identified. –Placebo expectancy manipulations: decreases in “pain matrix” regions1,2, increases in control regions, particularly rACC2,3,4

–Placebo analgesia brain-behavior correlations = between-subjects only.

–Event-related (cue-based) expectancy manipulations: modulation of pain matrix and striatal regions5,6

–Have not examined relationship between brain and pain reports.•For a brain region or pathway to mediate expectancy effects on reported pain, its activity must:

a) be influenced by expectancy.b) predict trial-by-trial changes in reported pain, even within a single level of noxious stimulation.c) statistically explain a significant portion of expectancy effects on trial-by-trial reported pain.

•We used multi-level mediation (M3) software to test this compound hypothesis, and to locate regions that formally mediate the relationship between experimentally manipulated expectancy and reported pain.

Multiple mediators of expectancy effects on pain perception:Interactions among higher-order brain regions and the pain matrix

Lauren Y. Atlas1, Niall Bolger1, Martin Lindquist2, Tor D. Wager1

1Columbia University Department of Psychology, 2Columbia University Department of Statistics

INTRODUCTION

METHODS

RESULTS

•Voxelwise single trial analysis

•Whole-brain multi-level mediation

http://www.scan.psych.columbia.edu/Columbia Psychology SCAN group

Three linear equations:

1. y = cx + ey

2. m = ax + em

3. y = bm + c'x + e'y

•Experimental design

A test for mediation should satisfy the following criteria:

1. M should be related to X (a effect)2. M should be related to Y after

controlling for X (b effect)

3. The indirect relationship (a*b)should be significant

Contact: laurenatlas@gmail.com

•18 subjects (mean age = 25.3, 9 females)•Thermal stimulation to 4 sites on left forearm with 16mm thermode (Medoc, Inc.))•FMRI acquisition: 1.5T GE scanner, Spiral In-Out, 29 slices, 2s TR

3. Two conditioning runs + fMRI

4. Six experimental runs + fMRI

1. Pain calibration

Low pain cueLow pain cue

Low Pain (Level 2) LLMed Pain (Level 5) LM

High pain cueHigh pain cue

Med Pain (Level 5) HMHigh Pain (Level 8) HH

2. Learning Task: Tones predict low or high pain, counterbalanced across subjects; S informed of cue-pain pairings, must correctly identify 90% of tones to proceed

Level 2 (Slightly painful)

Level 5 (Moderate pain)

Level 8 (Max tolerable pain)

++

++

CUEDelay

HEATISI

2s6s

10s14s 4s

8sHow Painful?

ISI

+

Contrast HM vs LM during pain period: identical stimulation, high vs low pain expectancy

1. Fit basis functions trial-by-trial

2. Estimate trial-by-trial area under the curve (AUC)

3. Use trial level parameters in multi-level mediation

324 Schermerhorn HallDepartment of Psychology1190 Amsterdam Ave.New York, NY 10027

Basis functions Trial level parameters

Trial-by-trial estimates

•HypothesesI. Expectancies modulate pain reports.II.Expectancies modulate pain matrix activity (Path

a).III.Pain matrix activity affects perceived pain (Path

b).IV.Expectancy effects in pain matrix regions lead to

changes in perceived pain (Mediation effect, A*B).V. Higher order regions involved in cognitive control

and value processing also mediate expectancy effects on perceived pain.

VI.Interactions among these mediators predict perceived pain.

Perceived Pain

42.52 (6.18)***

2.24 (0.77)**6.25 (1.72) *** 0.20 (0.08)**Expectancy (HM-LM)

Expectancy (HM-LM)

Perceived Pain

dACC

R. Thalamus

CB VermisPons

L. Anterior Insula

IV. Pain matrix mediators VI. Secondary mediation analysis: Expectancy-based interactions between pain matrix and higher-order mediators contribute to perceived pain.

SUMMARY•Nearly all pain matrix regions are modulated by expectancy during noxious thermal stimulation (all except dACC, PAG).•Some pain matrix regions (Insula, dACC) predict perceived pain controlling for expectancy and temperature.•A key subset of pain matrix regions (insula, thalamus, dACC, pons, cerebellum) mediate expectancy effects on perceived pain. •Amygdala, striatum, and medial OFC mediate expectancy effects on pain matrix mediators.•R. dmPFC mediates expectancy effects on pain independent of pain matrix activity.

Pain Localizer - HH>LL

YNYR. Thalamus*

YNYPons*NNNPAGYYNdACC*

NNYL. Middle Insula

NNYR. Middle Insula

YYYL. Anterior Insula

NYYR. Anterior Insula†

NNYR. SIIYYYCB Vermis

NYYBilateral Cerebellum†

AB?B?A?Region

WHOLE BRAIN MULTI-LEVEL MEDIATION ANALYSIS

1. DD Price et al. Pain, 127, 63-72 (2007).2. TD Wager et al. Science, 303, 1162-7

(2004).3. U Bingel et al. Pain, 120, 8-15 (2006).4. P Petrovic, E Kalso, KM Petersson, M

Ingvar. Science, 295, 1737-40, (2002).5. T Koyama, JG McHaffie, PJ Laurienti, RC

Coghill. PNAS, 102, 12950-12955 (2005).6. JR Keltner et al. JNeuroscience, 26, 4437-

4443 (2006).

Pain

Rati

ng

Pain

Rating

Reprints: http://www.columbia.edu/cu/psychology/tor/

I. Expectancies modulate perceived pain. HM>LM (p<.0001), allowing us to examine mediators of expectancy effects on reported pain within a single level of noxious stimulation.

0

1

2

3

4

5

6

7

8

LL LM HM HH

Perceived Pain

****

Perceived Pain (Pain Reports)

Brain Mediators

Expectancy (HM-LM)

A*B: Which brain regions mediate expectancy effects on pain perception?

Path bPath a

Path c/c’

Pons

II. Path a: Expectancies modulate brain activity during pain.

III. Path b: Controlling for expectancy and temperature, brain activity predicts perceived pain.

LM HM

LM HM

PAIN PERIOD ACTIVATION EXPECTANCY

Medial OFC Right VLPFCLeft DLPFC Left Amygdala

Path a

HE

-Med

> L

E-M

ed

PonsRight ThalamusRight SII Left Anterior Insula

Pai

n R

egio

nsa.

b.

LE-M

ed >

HE

-Med

Medial OFCRight DMPFC Right IFG

Pos

itive

effe

ct

Right latPFC

rdACCPai

n R

egio

ns

a.

b.

Neg

ativ

e ef

fect

PAIN PERIOD ACTIVATION PERCEIVED PAIN Path b

1

5

4

3

2

Rep

orte

d P

ain

Mid. Cingulate

1

5

4

3

2

Rep

orte

d P

ain

R. Ant. Insula

1

5

4

3

2

Rep

orte

d P

ain

Cerebellum

1

5

4

3

2

Rep

orte

d P

ain

IV, V. Mediation effect: Pain matrix regions and higher order regions mediate expectancy

effects on reported pain.

L. Ant. Insula

L (Ipsilateral) R (Contralateral)

Vermis

L. Putamen, Caudate

rdACCpgACC

L. DLPFCTPJ

R. Thalamus

Mid. CingdACC

Pons

Middle Occipital Gyrus

R. dmPFC

R sgACC

References

Mediating expectancy effects on pain matrix

Expectancy Perceived Pain

Pain Matrix Mediators

Left Amygdala

Left Ventral Striatum

SII

Medial OFC

Expectancy Perceived Pain

Pain Matrix Mediators

Right dmPFC

-6.3994 (2.44)** 0.2252 (.05)**

Expectancy (HM-LM)

Perceived Pain

Pain Matrix Mediators*

Expectancy Region

ab

Path c

a

b

Path c

*:Activity averaged across pain matrix mediators.

Thick arrows depict paths identified in primary mediation (Hypoth IV). Dashed arrows depict effects potentially being mediated.

Report Region

Independent of pain matrix

RegionL VSL AmygR SIImOFC

A*B0.4333(.21)*0.5039 (.25)*0.3749(.11)**0.2468(.18)*

Secondary mediation 1Secondary mediation 2

Positive path coefficientsNegative path coefficients