Trajectories for ADNI

The fixed effect “M*GENDER”, meaning that the effect of each month in predicting the values of the biomarkers was modulated by gender of the individual. This effect played a significant role for some of the biomarkers at a certain stage. These results are presented as follows:

  • Hippocampal volume in normal subjects was observed to be lower in females at each time point of measurement.

  • Ventricle volume in normal and demented subjects was observed to be higher in females.

  • CSF Aβ in normal subjects was found to be higher in males.

  • Plasma Aβ42/Aβ40 in MCI subjects was found to be lower in females.

Comparison of literature baseline cohort [1] (Victor L Villemagne cohort) with ADNI level baseline data:
Some of the examples are shown in Figures 14 and 15.

Age did not prove to be an effective parameter for the overall progression of the disease as there was a lack of available data to support this hypothesis. The “stage” of the disease was observed to be a significant parameter as some patterns were identified for the biomarkers. The parameter, “repeated measurements at different time points”, proved to be the best, as we could see an informative pattern for different biomarkers in a longitudinal fashion. This hypothesis was also backed by the updated hypothetical model of Jack, et al.

Thus, the algorithm underlying the longitudinal model was developed considering “repeated measurements at time (months)” and stage of the diseases as essential parameters. Linear mixed modelling approach facilitated the prediction of the response values of the biomarker for each individual by taking into account varied predictors. The predictors were tested for their significance and some of these predictors exhibited a high significant role in the prediction of biomarkers values. One of the predictors having a significant role was “gender” of each subject.

The trajectories for volumetric measures and the cognitive measures in the longitudinal viewer depicted a clear progression through various stages of AD. The frequency of subjects decreases along the progression of the disease. Aβ shows a sudden increase while transitioning from MCI to Dementia stage. This unusual pattern can be due to the fact of increase in the number of subjects at that particular month.

The re-sampling approach adapted by us to structure the data available in the literature also proved to be significant, as it enabled a valuable comparison of literature to the ADNI data. The pattern of biomarkers along the disease progression in both the data sets were observed to be similar. However, the algorithm underlying this approach needs to be made more robust so that it could be extended to the diverse set of literature data available.

The model developed helps to predict the value of the biomarker from diverse set of predictors, so it can be advantageous for a data set where a particular biomarker is missing. For example, if we do not have ABETA values for a particular stage but we have its predictors, we can predict the ABETA value for the different time points of measurement and hence its trajectory for the particular stage.

The model allows us to observe the likely progression of the disease in the normal population as the longitudinal model is predicted over distinct groups of population in separate stages of AD.

Limitations of the model

  • The model is limited to the set of biomarkers, which have been discussed in this report. Although, availability of further biomarker data could enable the model to make predictions on them as well.

As the trajectories obtained are generated from the predicted mean values of each biomarker, it could lack specificity for individual subjects.

Age did not prove to be an effective parameter for the overall progression of the disease as there was a lack of available data to support this hypothesis. The “stage” of the disease was observed to be a significant parameter as some patterns were identified for the biomarkers. The parameter, “repeated measurements at different time points”, proved to be the best, as we could see an informative pattern for different biomarkers in a longitudinal fashion. This hypothesis was also backed by the updated hypothetical model of Jack, et al.

Thus, the algorithm underlying the longitudinal model was developed considering “repeated measurements at time (months)” and stage of the diseases as essential parameters. Linear mixed modelling approach facilitated the prediction of the response values of the biomarker for each individual by taking into account varied predictors. The predictors were tested for their significance and some of these predictors exhibited a high significant role in the prediction of biomarkers values. One of the predictors having a significant role was “gender” of each subject.

The trajectories for volumetric measures and the cognitive measures in the longitudinal viewer depicted a clear progression through various stages of AD. The frequency of subjects decreases along the progression of the disease. Aβ shows a sudden increase while transitioning from MCI to Dementia stage. This unusual pattern can be due to the fact of increase in the number of subjects at that particular month.

The re-sampling approach adapted by us to structure the data available in the literature also proved to be significant, as it enabled a valuable comparison of literature to the ADNI data. The pattern of biomarkers along the disease progression in both the data sets were observed to be similar. However, the algorithm underlying this approach needs to be made more robust so that it could be extended to the diverse set of literature data available.

The model developed helps to predict the value of the biomarker from diverse set of predictors, so it can be advantageous for a data set where a particular biomarker is missing. For example, if we do not have ABETA values for a particular stage but we have its predictors, we can predict the ABETA value for the different time points of measurement and hence its trajectory for the particular stage.

The model allows us to observe the likely progression of the disease in the normal population as the longitudinal model is predicted over distinct groups of population in separate stages of AD.

Limitations of the model

  • The model is limited to the set of biomarkers, which have been discussed in this report. Although, availability of further biomarker data could enable the model to make predictions on them as well.

As the trajectories obtained are generated from the predicted mean values of each biomarker, it could lack specificity for individual subjects.

Volumetric measures


Fig_1Figure 1: Hippocampal Volume progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: : Transformed values of hippocampal brain volume, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Fig_2Figure 2: Entorhinal Volume progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: : Transformed values of entorhinal brain volume, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Fig_3Figure 3: Ventricle Volume progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: : Transformed values of ventricle brain volume, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Fig_4Figure 4: Inracranial Volume (ICV) progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: : Transformed values of intra cranial brain volume, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

CSF measures


Figure 5: CSF ABETA progression via various stages of AD, x-axis: Time points of measurement (in months 0, 24, 36, 48, 60) for normal subjects and time points of measurement (in months 0, 12, 36, 48) for MCI stage; scaled from 0 to 1; Upper figure: y-axis: : Transformed values of CSF ABETA, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Figure 6: CSF PTAU progression via various stages of AD,
x-axis: Time points of measurement (in months 0, 24, 36, 48, 60) for normal subjects and time points of measurement (in months 0, 12, 36, 48) for MCI stage; scaled from 0 to 1; Upper figure: y-axis: : Transformed values of CSF PTAU, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Figure 7: CSF TAU progression via various stages of AD, x-axis: Time points of measurement (in months 0, 24, 36, 48, 60) for normal subjects and time points of measurement (in months 0, 12, 36, 48) for MCI stage; scaled from 0 to 1; Upper figure: y-axis: : Transformed values of CSF TAU, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Cognitive tests


Figure 8: CDR-SB progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: : Range of values of CDRSB ranging from 0-18, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Figure 9: Cognition measure progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: : Range of values of cognition score (mix of MOCA and MMSE, ranging from 0-30), Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Figure 10: MMSE progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: : Range of values of MMSE, ranging from 0-30, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Plasma measures


Figure 11: Plasma Abeta42 to Abeta40 ratio; progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: Values of biomarker plasma Abeta42RatioAbeta40, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Figure 12: Plasma Abeta40 progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: : Values of biomarker plasma Abeta40, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

Figure 13: Plasma Abeta42 progression via various stages of AD, x-axis: Time points of measurement (in months 0, 6, 12, 24) for each stage scaled from 0 to 1; Upper figure: y-axis: Values of biomarker plasma Abeta42, Lower figure: y-axis: Number of subjects measured at different months (considered for each stage)

MMSE pattern comparison


Figure14: Comparison of MMSE progression in three stages of AD (“NL”, “MCI”, “Dementia”) between left( ADNI baseline) and right (cohort generated from literature) illustrated in the form of Boxplot: x-axis: Normal (NL (rust color), MCI (green color) and Dementia stag (blue color) of the disease, y-axis: MMSE scale ranging from 17.5 to 30.
[View full image]

CDR-SB pattern comparison


Figure 15: Comparison of CDR-SB progression in three stages of AD (“NL”, “MCI”, “Dementia”) between left (ADNI baseline) and right (cohort generated from literature)illustrated in the form of Boxplot: x-axis: Normal (NL (rust color), MCI (green color) and Dementia stag (blue color) of the disease, y-axis: MMSE scale ranging from 0 to 10.
[View full image]