3.1 Demographic Analysis
CR and LR subjects differed with respect to age (F1,30=5.48, p=.03) and educational attainment (F1,29=8.44, p=.01), but not parental education (Wilks’ F2,24=1.44, p=.26) or sex (χ2=3.14, p=.08). Both groups had similar smoking habits as quantified by packs per day (F1,28=0.84, p=.37). Demographic and clinical information are presented in .
Demographic/Clinical Characteristics and Raw Odor Detection Thresholds
3.2 Odor Detection Threshold Performance
There was a statistically significant interaction of group × odorant (F1,30=4.05, p=.05) (). In paired contrasts, CR subjects were significantly worse than LR subjects at detecting lyral (F1,30=6.42, p=.02, Cohen’s d=0.90), but not citralva (F1,30>0.01, p>.99, Cohen’s d<0.01). Inclusion of age and sex as covariates did not alter these observed effects, although age emerged as a significant independent predictor of performance (age: F1,29=4.71, p=.04, sex: F1,29=0.04, p=.84). Exclusion of the 3 CR subjects who had a schizophrenia family member did not alter the results (group × odorant: F1,27=7.99, p<.01).
Mean odor threshold detection sensitivities to citralva and lyral in youth at clinical and low risk for psychosis
3.3 Odor Identification and Discrimination Performance
There was a significant group difference (F1,29
<.001), with CR subjects performing worse than LR subjects (see ). There was also a significant test effect (F1,29
<.01); scores on odor discrimination were lower, overall, than odor identification scores. However, there was no group × test interaction (F1,29
=.55) and the groups differed significantly on each of the individual tests (identification: F1,29
<.001; discrimination: F1,29
=.01). Covarying for age and sex did not alter these results; this replicates our previous findings in a smaller sample (Kamath et al., in press
3.4 Relationships Among Olfactory Performance Measures
We considered whether the measures on which CR subjects were impaired (lyral detection threshold, odor discrimination, odor identification) were correlated, suggesting disruption of a common underlying olfactory substrate - or whether they were relatively independent, suggesting multiple discrete abnormalities. There was a significant correlation, across groups, between odor identification and odor discrimination (r =.41, p=.02). However, these measures were not significantly correlated with either lyral detection threshold (identification: r = −.30, p=.11; discrimination: r = −.01; p=.98) or citralva detection threshold (identification: r = −.06, p=.73; discrimination: r = .18; p=.33). There was also no relationship between the detection thresholds for lyral and citralva (r = .27; p=.14). Examination of just the CR group revealed a similar pattern, though with lower significance reflecting the smaller sample size (identification vs. discrimination: r =.47, p=.07; lyral threshold vs. identification: r =.15, p=.57; lyral threshold vs. discrimination: r =.22, p=.42; lyral threshold vs. citralva threshold: r =.31, p=.25; citralva threshold vs. identification: r =−.26, p=.33; citralva threshold vs. discrimination: r =.35, p=.18). This suggests that odor identification and discrimination abnormalities may share a common underlying olfactory neuropathology, but abnormal lyral threshold detection may denote a separate pathological substrate. Moreover, the absence of any association between the threshold detections for lyral and citralva indicates that this is not just a generic dissociation of odor detection sensitivity from other higher-order olfactory domains.
3.5 Sensitivity and Specificity of Olfactory Measures
The ROC curves for each of the individual test scores and the omnibus measure are presented in . Consistent with the findings above, AUC exceeded 0.5 (indicating better-than-chance predictive capacity) for each individual test except citralva threshold detection. AUC was 0.748±0.092 SE for lyral threshold, 0.894±0.052 SE for odor identification, and 0.758±.087 SE for odor discrimination. These values indicate the proportion of subjects correctly classified as CR or LR, based solely on performance on each individual test. Consistent with the idea that these are not redundant tests, but rather reflect relatively independent areas of impairment, the AUC for the omnibus score was substantially higher, 0.971±0.027 SE, indicating near-perfect classification accuracy when these olfactory measures are used in combination.
ROC Curve Analysis of Olfactory Tasks in Clinical Risk and Low Risk Subjects
3.5 Relationships Between Olfactory Performance and Clinical Symptoms
None of the individual olfactory measures or the composite score was associated with prodromal symptom severity, as measured by the SOPS total score (all p>0.20).