Table 7 Percent of variance in morning plasma AD biomarker levels explained by the neuro-glymphatic model over the null modela

From: The glymphatic system clears amyloid beta and tau from brain to plasma in humans

Condition and Predictor

Amyloid Status

Plasma Biomarker

Variance Explained

CI

P

Sleep Condition

PhysioS

Positive

All Biomarkers

51.9%

38.6% – 84.3%

0.005

 Ab40

49.1%

22.4% – 92.8%

0.027

 Ab42

56.0%

5.4% – 91.4%

0.043

 All Tau

51.7%

35.6% – 83.7%

0.005

Negative

All Biomarkers

91.1%

47.4% – 98.1%

0.013

 Ab40

97.8%

45.5% – 99.6%

0.015

 Ab42

93.7%

40.7% – 99.3%

0.017

 np-tau217

95.4%

42.9% – 99.3%

0.017

 np-tau181

94.1%

38.7% – 99.1%

0.018

 p-tau181

71.3%

49.0% – 97.3%

0.007

Sleep Condition

HypnoS

Positive

All Biomarkers

42.0%

32.5% – 81.8%

0.006

 Ab40

57.7%

16.1% – 91.4%

0.032

 Ab42

62.5%

9.1% – 90.8%

0.032

 All Tau

38.4%

28.9% – 80.7%

0.006

Negative

All Biomarkers

87.4%

17.7% – 96.9%

0.020

 Ab40

95.7%

-30.1% – 98.3%

0.081

 Ab42

87.1%

-39.7% – 97.2%

0.134

 np-tau217

93.1%

21.4% – 98.6%

0.029

 np-tau181

90.9%

1.5% – 98.4%

0.059

 p-tau181

64.0%

26.4% – 96.8%

0.009

Sleep Deprivation/Wake

PhysioW

Positive

All Biomarkers

68.9%

65.5% – 92.5%

0.018

 Ab40

76.6%

-7.9% – 94.1%

0.092

 Ab42

-53.8%

−13.7% – 95.4%

0.073

 All Tau

72.8%

68.0% – 92.7%

0.018

Negative

All Biomarkers

85.1%

34.9% – 98.0%

0.020

 Ab40

92.5%

40.2% – 99.4%

0.019

 Ab42

93.3%

44.7% – 99.5%

0.021

 np-tau217

60.7%

−184.8% – 95.9%

0.372

 np-tau181

84.3%

−44.6% – 98.7%

0.098

 p-tau181

73.8%

34.9%96.8%

0.013

  1. aCI, 95% confidence interval; P, p-value; PhysioS and PhysioW, single predictors constructed from the linear combination of the full model predictors during the sleep and sleep deprivation (wake) conditions, respectively, for both amyloid positive and negative participants separately; HypnoS, single predictor constructed from the full model hypnogram features during the sleep condition for both amyloid positive and negative participants separately. Confidence intervals were calculated using the 95% quantiles of 1000 parametric bootstrap replicates. Two-sided p-values were obtained from the parametric bootstrap by calculating the proportion of bootstrap estimates that fell below the null hypothesis of zero explained variance and multiplying this proportion by two. P-values were not adjusted for multiple comparisons. Bolded confidence intervals and P values are statistically significant at the 0.05 level.