Periodic Reporting for period 4 - TRANS-ID (TIME FOR A CHANGE: An out-of-the-box approach to unravel transitions in depression)
Période du rapport: 2021-02-01 au 2021-07-31
Depression and other forms of psychopathology are highly prevalent. There is an urgent need for improved insight in psychopathology and for accurate personalized risk assessment. The general aim of the TRANS-ID project was to examine whether psychopathological conditions behave according to principles of complex systems and, specifically, whether early warning signals (EWS) can be detected anticipating major and sudden shifts in psychopathology. We examined presence of EWS before individuals experience sudden transitions towards higher levels of depressive symptoms (in individuals tapering their antidepressants), towards lower levels of depressive symptoms (in individuals with depression starting psychological treatment), and towards drops in mental health (in young individuals at heightened risk for psychopathology). We examined EWS in affect, symptom, and physiological data.
The primary outcomes of the TRANS-ID studies provide important empirical insight in the phenomenon of Critical Slowing Down (CSD) as an EWS of upcoming symptom shifts at the idiographic level. The findings indicate that EWS are not as generic as hypothesized. There seem to be individual differences in which mental states and physiological measures showed EWS. The fact that EWS were found more often in individuals with transitions compared to individuals without transitions supports the existence of the phenomenon of CSD before symptom transitions. However, EWS did not precede transitions consistently enough to be promising as a clinical tool for monitoring the risk of upcoming symptom transitions.
The primary outcomes of the TRANS-ID studies provide important empirical insight in the phenomenon of Critical Slowing Down (CSD) as an EWS of upcoming symptom shifts at the idiographic level. The findings indicate that EWS are not as generic as hypothesized. There seem to be individual differences in which mental states and physiological measures showed EWS. The fact that EWS were found more often in individuals with transitions compared to individuals without transitions supports the existence of the phenomenon of CSD before symptom transitions. However, EWS did not precede transitions consistently enough to be promising as a clinical tool for monitoring the risk of upcoming symptom transitions.
Work performed
We carried out challenging time-series studies with an extremely large numbers of observations per individual precisely during those periods in which symptom transitions occurred. We have recruited the aimed number of participants and achieved high compliance rates. We have developed novel sampling procedures and tested novel instruments to optimally measure heart rate variability. We have pre-processed the data. New methods were tested, statistical procedures have been optimized, and new analytical skills have been acquired.
Main results
First, we tested whether psychopathology behaves according to principles of dynamic systems in different data-sets. In a first study, we found limited evidence that EWS inform on the type of symptoms that will undergo a transition (Schreuder et al., 2020). In another study, we showed that symptom domains did not become more interrelated and formed more clusters as illness severity increased (Groen et al., 2019). In a confirmatory single-subject study, we showed that EWS significantly increased before recurrence of depressive symptoms in a patient who discontinued antidepressants (Wichers, Smit & Snippe , 2020). In a study in bipolar patients, we found that actigraphy data derived EWS and spectral analysis show promise for the prediction of upcoming mood transitions (Kunkels et al, 2021).
Second, we examined EWS in formerly depressed patients who discontinued their antidepressants (WP1). Results showed that rises in autocorrelation and variance in affect preceded transitions towards higher levels of depressive symptoms in a part of the sample. These EWS were more common patients with a transition (~32.9%) compared to patients without a transition (~16.2%). The percentage of true positives was higher (52.6%) when examining whether EWS were found in at least 3 affect measures, while at the same time reducing the percentage of false positives (11.1%).
Third, we examined EWS in a sample of depressed individuals who started psychological treatment (WP2). Results showed that rises in autocorrelation and variance in affect preceded transitions lower levels of depressive symptoms again in a part of the sample (~44%), but also in part of the individuals without a transition, although in fewer individuals (~27%). The percentage true positives was higher when examining EWS in one of the variables (~89%), however the percentage of false positives was also higher (~62.5%).
Forth, we examined actigraphy derived EWS (e.g. kurtosis, autocorrelation) in formerly depressed patients who discontinued their antidepressants (WP1/WP3). Preliminary results showed that EWS preceded transitions in eight out of sixteen individuals (50.0%), while generating false positives in three out of nine participants (22.2%). The strongest EWS was kurtosis, which preceded transitions in 4 out of sixteen individuals (25.0%), while generating a false positive in one out of nine participants (11.1%).
Fifth, the complexity and variability of heart rate dynamics in formerly depressed patients who discontinued their antidepressants (WP1/WP3) was tested. Preliminary, results indicate that the variability and complexity measures have high predictive value in distinguishing between individuals with recurrence of depression and those who stay in remission.
Sixth, we showed that the predictive value of EWS was low in a sample of at-risk youth (WP4). Rising autocorrelations in specific mental states seldomly preceded drops in mental health (~5% sensitivity), and also seldomly occurred in individuals without a drop in mental health (~96 specificity). Although a rise in autocorrelation in at least one of the mental states was found for the majority of the individuals preceding drops in mental health (70%), this was also found for the majority of the individuals without drops in mental health (65%).
Seventh, we showed that a simpler measure, namely structural mean changes, detected using Statistical Process Control (SPC) charts, may hold promise as a tool for clinical practice to detect upcoming transitions towards higher levels of depressive symptoms: the accuracy of this method was higher than of EWS and change can be detected in real-time.
Exploitation and Dissemination
We have so far published 40 scientific papers in peer-reviewed journals, mostly in top 25 journals. Another 15 scientific papers are submitted or in preparation. Oral presentations of TRANS-ID team members were given at 14 international conferences, at 6 conferences for clinical audiences, and 9 scientific expert meetings in the Netherlands. We organized a workshop in collaboration with Prof. Patrick McGorry for key players from the international research community. TRANS-ID research was communicated in national newspapers and radio broadcasts. Additional activities involved the development of an R package (ACTman) and a collaboration with the data management team (Roqua) and the iLab of the psychiatry department UMCG to develop a flexible interface for personalized monitoring.
We carried out challenging time-series studies with an extremely large numbers of observations per individual precisely during those periods in which symptom transitions occurred. We have recruited the aimed number of participants and achieved high compliance rates. We have developed novel sampling procedures and tested novel instruments to optimally measure heart rate variability. We have pre-processed the data. New methods were tested, statistical procedures have been optimized, and new analytical skills have been acquired.
Main results
First, we tested whether psychopathology behaves according to principles of dynamic systems in different data-sets. In a first study, we found limited evidence that EWS inform on the type of symptoms that will undergo a transition (Schreuder et al., 2020). In another study, we showed that symptom domains did not become more interrelated and formed more clusters as illness severity increased (Groen et al., 2019). In a confirmatory single-subject study, we showed that EWS significantly increased before recurrence of depressive symptoms in a patient who discontinued antidepressants (Wichers, Smit & Snippe , 2020). In a study in bipolar patients, we found that actigraphy data derived EWS and spectral analysis show promise for the prediction of upcoming mood transitions (Kunkels et al, 2021).
Second, we examined EWS in formerly depressed patients who discontinued their antidepressants (WP1). Results showed that rises in autocorrelation and variance in affect preceded transitions towards higher levels of depressive symptoms in a part of the sample. These EWS were more common patients with a transition (~32.9%) compared to patients without a transition (~16.2%). The percentage of true positives was higher (52.6%) when examining whether EWS were found in at least 3 affect measures, while at the same time reducing the percentage of false positives (11.1%).
Third, we examined EWS in a sample of depressed individuals who started psychological treatment (WP2). Results showed that rises in autocorrelation and variance in affect preceded transitions lower levels of depressive symptoms again in a part of the sample (~44%), but also in part of the individuals without a transition, although in fewer individuals (~27%). The percentage true positives was higher when examining EWS in one of the variables (~89%), however the percentage of false positives was also higher (~62.5%).
Forth, we examined actigraphy derived EWS (e.g. kurtosis, autocorrelation) in formerly depressed patients who discontinued their antidepressants (WP1/WP3). Preliminary results showed that EWS preceded transitions in eight out of sixteen individuals (50.0%), while generating false positives in three out of nine participants (22.2%). The strongest EWS was kurtosis, which preceded transitions in 4 out of sixteen individuals (25.0%), while generating a false positive in one out of nine participants (11.1%).
Fifth, the complexity and variability of heart rate dynamics in formerly depressed patients who discontinued their antidepressants (WP1/WP3) was tested. Preliminary, results indicate that the variability and complexity measures have high predictive value in distinguishing between individuals with recurrence of depression and those who stay in remission.
Sixth, we showed that the predictive value of EWS was low in a sample of at-risk youth (WP4). Rising autocorrelations in specific mental states seldomly preceded drops in mental health (~5% sensitivity), and also seldomly occurred in individuals without a drop in mental health (~96 specificity). Although a rise in autocorrelation in at least one of the mental states was found for the majority of the individuals preceding drops in mental health (70%), this was also found for the majority of the individuals without drops in mental health (65%).
Seventh, we showed that a simpler measure, namely structural mean changes, detected using Statistical Process Control (SPC) charts, may hold promise as a tool for clinical practice to detect upcoming transitions towards higher levels of depressive symptoms: the accuracy of this method was higher than of EWS and change can be detected in real-time.
Exploitation and Dissemination
We have so far published 40 scientific papers in peer-reviewed journals, mostly in top 25 journals. Another 15 scientific papers are submitted or in preparation. Oral presentations of TRANS-ID team members were given at 14 international conferences, at 6 conferences for clinical audiences, and 9 scientific expert meetings in the Netherlands. We organized a workshop in collaboration with Prof. Patrick McGorry for key players from the international research community. TRANS-ID research was communicated in national newspapers and radio broadcasts. Additional activities involved the development of an R package (ACTman) and a collaboration with the data management team (Roqua) and the iLab of the psychiatry department UMCG to develop a flexible interface for personalized monitoring.
The TRANS-ID project is the first to test systematically, empirically and in repeated experiments whether psychopathology behaves according to principles of complex systems in terms of the presence of early warning signals. The project thereby makes an important step forward in the application of complex system theory to the field of psychiatry.
Second, the TRANS-ID project has delivered datasets with fine-grained data over extended periods of time that directly precede symptom transitions. Such datasets give the field the opportunity to examine what happens in individuals before transitions.
Third, the TRANS-ID project contributed strongly to the paradigm shift from group-based towards personalized analyses.
Second, the TRANS-ID project has delivered datasets with fine-grained data over extended periods of time that directly precede symptom transitions. Such datasets give the field the opportunity to examine what happens in individuals before transitions.
Third, the TRANS-ID project contributed strongly to the paradigm shift from group-based towards personalized analyses.