Skip to main content

Table 2 Biomarker properties. For extended data see Supplemental Tables.

From: The Autism Biomarkers Consortium for Clinical Trials: evaluation of a battery of candidate eye-tracking biomarkers for use in autism clinical trials

 

OMI

AM

SI

SS

BM

PLR

Signal Acquisition and Validity at T1

ASD

 Acquisition valid

279 (99%)

280 (100%)

280 (100%)

279 (100%)

280 (100%)

278 (99%)

 Signal valid

272 (97%)

280 (100%)

276 (99%)

273 (98%)

277 (99%)

266 (95%)

 Site differences

X2 = 6.6, p = .16

–

X2 = 1.0, p = .90

X2 = 7.6, p = .11

X2 = 5.2, p = .27

X2 = 16.1, p < .01

TD

 Acquisition valid

119 (100%)

119 (100%)

119 (100%)

119 (100%)

119 (100%)

119 (100%)

 Signal valid

119 (100%)

119 (100%)

119 (100%)

119 (100%)

119 (100%)

117 (98%)

 Site differences

–

–

–

–

–

X2 = 3.0, p = .56

Task

OMI

AM

SI

SS

BM

PLR

Construct validity for each ET task based on TD participants at T1

Construct

Face preference

Face preference

Face preference

Face preference

Biomotion preference

Pupillary light reflex

Test

All sub-tasks valid

> Chance face gaze

> Chance face gaze

> Chance face gaze

> Chance biomotion gaze

Pupil constriction post flash

Null hypothesis

–

%Face = 3.2%

%Face = 8.3%

%Face = 3.9%

%Bio = 50%

Constrict > 0

Sample values

AM✓ SI✓ SS✓

M = 27.6%, SD = 8.5%

M = 30.4%, SD = 9.6%

M = 34.9%, SD = 8.0%

M = 54.8%, SD = 6.1%

M = .505, SD = .074

Statistic

–

t(118) = 31.3

t(118) = 25.0

t(118) = 42.1

t(118) = 8.5

t(116) = 73.9

p

–

< .001

< .001

< .001

< .001

< .001

d

–

2.87

2.30

3.88

.79

6.83

Validity

✓

✓

✓

✓

✓

✓

Group

df range

OMI

AM %Face

SI %Face

SS %Face

BM %Bio

PLR latency

Six-week stability (ICC) from T1 to T2 for primary pre-specified biomarkers

TD

106 ≤ df ≤ 112

.828

.848

.842

.569

.441

.866

ASD

223 ≤ df ≤ 274

.836

.834

.813

.680

.505

.749

ASD < 8.5 years

116 ≤ df ≤ 144

.837

.820

.813

.657

.462

.775

ASD ≥ 8.5 years

106 ≤ df ≤ 129

.832

.842

.812

.695

.530

.725

ASD IQ < 75

24 ≤ df ≤ 30

.754

.801

.794

.476

.238

.741

ASD IQ ≥ 75

198 ≤ df ≤ 243

.849

.836

.819

.701

.526

.749

 

OMI

AM %Face

SI %Face

SS %Face

BM %Bio

PLR latency

Group discrimination at T1 for primary biomarkers

M (SD)

 ASD values

24.4% (8.5%)

18.7% (8.7%)

24.2% (10.7%)

29.9% (9.8%)

53.4% (6.8%)

285 ms (15 ms)

 TD Values

30.9% (7.6%)

27.6% (8.5%)

30.4% (9.6%)

34.9% (8.0%)

54.8% (6.1%)

279 ms (15 ms)

No covariates

 Statistic

F(1,389) = 55.6

F(1,397) = 90.7

F(1,393) = 31.5

F(1,390) = 27.7

F(1,394) = 4.0

F(1,381) = 10.1

 p

< .001

< .001

< .001

< .001

.046

.002

 d

− .788

− 1.037

− .593

− .537

− .211

.350

 Î·p2

.117

.184

.069

.058

.009

.025

Age + IQ + Acq + Site Control

 Statistic

F(1,382) = 24.5

F(1,390) = 55.3

F(1,386) = 21.8

F(1,383) = 3.9

F(1,387) = 1.9

F(1,374) = 6.7

 p

< .001

< .001

< .001

.048

.165

.010

 Î·p2

.057

.120

.050

.009

.006

.018

Phenotypic characteristic

OMI

AM %Face

SI %Face

SS %Face

BM %Bio

PLR Latency

Spearman’s Correlations between ET and child behaviors in the ASD group at T1

Age

.115

.147*

.077

.087

− .063

.153*

Full IQ

.123*

.140*

.033

.192**

− .063

− .006

Verbal IQ

.183**

.188**

.122*

.199***

− .058

.026

NV IQ

.059

.081

− .029

.147*

− .072

− .030

ADOS SA

− .165**

− .228***

− .110

− .168**

.041

− .018

ADOS RRB

− .035

− .087

.001

.004

.024

− .100

VABS3 Soc SS

.125*

.150*

.103

.118

.008

− .032

VABS3 Com SS

.182**

.205***

.144*

.203***

.029

− .049

SRS SCI T

− .070

− .059

− .090

− .062

− .117

.096

SRS RIRB T

− .108

− .090

− .119*

− .092

− .041

.086

PDDBI SocApp T

.066

.130*

.079

.038

.146*

− .058

PDDBI REPRIT T

− .238***

− .211***

− .210***

− .191**

− .001

.106

Face Memory SS

.316***

.359***

.256***

.301***

− .045

.051

  1. OMI —Oculomotor Index of Gaze to Human Faces; BM = Biological Motion Preference; PLR — Pupillary Light Reflex; AM = Activity Monitoring; SI—Social Interactive; SS = Static Scenes; [Construct Validity] Task —ET task; Construct —hypothesized construct under investigation; Test —how the construct is tested; Null Hypothesis —formal definition of the construct validity test; Sample Values —TD performance on null hypothesis variable at T1; [Signal Acquisition] Acquisition Valid —participant generated ET data for some portion of the assay; Signal Valid —valid signal for primary DV (meeting all quality control criteria for admission of data); Difference in Site Valid Signal Rates —Pearson’s Chi-Squared test for site differences in valid signal (consistent with Monte Carlo simulation and unable to be computed for 100% data validity); [Six-week Stability] ICC —Intraclass Correlation Coefficients (ICC3); DV — dependent variable; IQ —DAS Full Scale IQ; df — degrees of freedom across task DVs in calculation of ICCs; [Group Discrimination] DV —dependent variable; Age = participant age; IQ = Full Scale IQ; Acq = %Valid data collection rate; Site = data collection site; d = Cohen’s d; ηp2 = partial eta squared; [Phenotypic Characteristic Correlations] Full IQ = DAS Full Scale IQ; NV IQ = DAS Nonverbal IQ; ADOS SA — Autism Diagnostic Observation Schedule social affect comparison score; ADOS RRB —restricted interests and repetitive behavior comparison score; VABS3 Soc — Vineland Adaptive Behavior Scales adaptive behavior socialization standard score; VABS3 Com — communication standard score; SRS-2 SCI —Social Responsiveness Scale social communication and interaction T-score; SRS-2 RRB — restricted interest and repetitive behavior T-score; PDDBI SocApp T — Pervasive Developmental Disorders Behavior Inventory Social Approach Behaviors T-score; PDDBI REPRIT — Repetitive, Ritualistic, and Pragmatic Problems Composite T-score; Face Mem SS — NEPSY memory for faces subtask score. *p < .05; **p < .01; ***p < .001. Underlined cells are significant even after controlling for Age, Full Scale IQ, and %Valid Data. Italicized cells cannot be controlled for these variables due to collinearity