Table 3 Coefficients of regression models associated with diagnostic outcomes, adjusting for age.

Predictor	ASD Diagnosis	ID Diagnosis	FSIQ	ABC
	OR [95% CI]	OR [95% CI]	B [95% CI]	B [95% CI]
PCCB gene¹	0.26 [0.05, 1.18]	5.50 [1.21, 30.10]	−9.85 [−29.10, 9.40]	−4.12 [−21.49, 13.26]
Two loss-of-function alleles²	3.85 [0.76, 22.83]	NA³	−19.2 [−38.94, 0.54]	−18.98 [−36.36, −1.6]
1-¹³C-Propionate oxidation, 60-min, % (ln)	0.57 [0.25, 1.18]	0.04 [ < 0.001, 0.24]⁴	15.72 [9.69, 21.76]	15.9 [10.89, 20.9]
Propionylcarnitine, plasma, µmol/L (ln)	1.88 [0.41, 10.55]	2203.06 [32.13, 8436433.3]	−30.18 [−43.78, −16.59]	−23.46 [−38.45, −8.46]
Erythropoietin, serum, mIU/mL (ln)	28.11 [2.55, 1050.72]	115.39 [6.82, 10245.56]	−36.77 [−52.62, −20.92]	−25.12 [−43.23, −7.01]
FGF21, plasma, pg/mL (ln)	0.79 [0.43, 1.38]	2.01 [1.1, 4.31]	−8.52 [−15.28, −1.77]	−7.88 [−13.58, −2.18]
GDF15, plasma, pg/mL (ln)	1.51 [0.62, 3.99]	59.83 [6.59, 2743.82]	−22.23 [−30.01, −14.45]	−15.92 [−24.28,−7.57]
Glutamine, plasma, µmol/L (ln)	0.07 [0, 0.9]	0.01 [0, 0.18]	51.07 [26.09, 76.05]	31.78 [7.47, 56.1]
Glutamine, urine, nmol/mg creatinine (ln)	0.48 [0.1, 2.07]	0.03 [0, 0.26]	17.88 [3.75, 32.01]	11.23 [−5.68, 28.13]
Glycine, plasma, µmol/L (ln)	0.85 [0.21, 3.31]	0.69 [0.17, 2.57]	5.59 [−12.44, 23.63]	6.05 [−9.42, 21.52]
Total 2-methylcitrate, plasma, nmol/L (ln)	1.36 [0.46, 4.44]	91.81 [7.41, 7176.88]	−16.11 [−26.26, −5.95]	−11.23 [−23.03, 0.58]

¹Reference is PCCA gene
²Reference is < 2 loss-of-function alleles
³All individuals with two loss-of-function alleles (n = 10) had a diagnosis of ID, compared to 10 of 23 in those with fewer than two loss-of-function alleles. Due to quasi-complete separation, computation of meaningful statistics was not possible.
⁴Quasi-complete separation occurred because the distributions of propionate oxidation were almost non-overlapping between those with and without ID. Firth’s logistic regression was used for this analysis.
ASD Autism spectrum disorder, ID Intellectual disability, FSIQ Full scale intelligence quotient, ABC Vineland adaptive behavior composite. All models contained age as a covariate. Continuous predictors were natural-log transformed, indicated by (ln). OR > 1 indicates that increased value of variable is associated with increased odds of diagnosis; OR < 1 indicate that increased value of variable is associated with decreased odds of diagnosis. Test statistics and exact p-values are found in Supplementary Table S3.

Search