Towards a New Therapy-Success Definition for the EuResist Prediction - PowerPoint PPT Presentation

Towards a New Therapy-Success Definition for the EuResist Prediction Engine Alejandro Pironti Computational Biology and Applied Algorithmics Max-Planck-Institut für Informatik

Motivation • Therapy-success rates in resource-rich settings have steadily increased over time • Reason: Improved potency, drug-resistance profile, and tolerability of therapies • Prediction of initial therapy response has become less challenging • Dichotomization of therapies into successes and failures based on a single (or a few) viral-load measurements problematic Figure: Kaplan-Meier probabilities of therapy continuation stratified by therapy start year. EuResist therapies with baseline genotypes. 2

Challenges Turbulent first therapy years Looking at first therapy year only can lead to mislabeling of therapies. Predict whether the viral load will become suppressed at all? Figure: Example therapy viral-load trajectory from the EuResist Database 3

Challenges Lack of adherence Using only one (or a couple) of viral-load measurements not sufficiently discriminative of good and bad therapies. Predict time to viral-load rebound? Figure: Example therapy viral-load trajectory from the EuResist Database 4

Challenges Transient viral-load rebound Time to first viral-load rebound is inadequate in some cases. Predict area under viral-load curve? Figure: Example therapy viral-load trajectory from the EuResist Database 5

Challenges Differential viral-load monitoring intervals Due to differential viral-load monitoring intervals, area under viral-load trajectory is not comparable among therapies. Figure: Example viral load trajectories for a Therapy from the EuResist Database 6

Proposed Solution • Use quantitative measure of therapy success Figure: Example therapy viral-load trajectory 7

Proposed Solution • Use quantitative measure of therapy success • Organize therapy viral-load trajectory into semesters Figure: Example therapy viral-load trajectory 8

Proposed Solution • Use quantitative measure of therapy success • Organize therapy viral-load trajectory into semesters • Use mean viral load for each semester Figure: Example therapy viral-load trajectory 9

Proposed Solution • Use quantitative measure of therapy success • Organize therapy viral-load trajectory into semesters • Use mean viral load for each semester • Count number of semesters with mean viral load under some threshold (aviremic semesters) • Working threshold: 125 copies per milliliter of blood serum • To the right: 9 aviremic Figure: Example therapy viral-load trajectory semesters 10

Right Censoring of the Number of Aviremic Semesters • Intent-to-treat (ITT) criterion • On-treatment (OT) criterion labels a number of aviremic labels a number of aviremic semesters as censored if: semesters as censored if: – Therapy is still ongoing – Therapy is still ongoing – There are semesters – There are semesters without viral-load without viral-load measurements measurements – Therapy was interrupted while viral load was suppressed Methods for performing regression with right-censored data are required 11

Proof-of-Concept Analysis ITT criterion: 4,529 censored (33%) OT criterion: 8,658 censored (64%) • Dataset: – 11,394 therapies from EuResist database – 2,211 therapy-failure genotypes • Therapy-failure genotypes are included as therapies with uncensored zero aviremic semesters • Protease and reverse- transcriptase genotypes • Integrase-inhibitor-use history as a surrogate for integrase genotype Figure: Histogram of the numbers of aviremic semesters for dataset 12

Proof-of-Concept Analysis • Use following features for • Regression method: predicting the log number of – Linear support vector aviremic semesters: machines for right- – Drug compounds censored data – Protease and reverse- • Performance assessment: transcriptase genotypes – Patient-wise disjoint 10-fold – Integrase-inhibitor-use cross validation history – Drug-exposure scores for protease and reverse- transcriptase inhibitors • Separate analyses for ITT and OT criteria 13

Proof-of-Concept Analysis: Results Performance measure: OT criterion: Harrell’s concordance index • 833,619 (91,411) usable pairs • Inspired on receiver-operating- • 645,102 (64,996) concordant characteristic curves pairs • Compares pairs of pairs of • Concordance probability: 0.77 measured and predicted (0.01) numbers of aviremic ITT criterion: semesters • 1,084,045 (32951) usable pairs • Two pairs are either unusable, • 782,716 (31,806) concordant concordant, or discordant pairs m 1 < m 2 ⇒ p 1 < p 2 ? • Concordance probability: 0.72 (m i , p i ): pair of measured and (0.01) predicted number of aviremic semesters Numbers averaged (standard deviation) across cross-validation folds 14

Outlook • What do you think of the number of aviremic semesters as a measure for therapy success? • Perform detailed analysis including test set • Include further variables as predictors: – Gender – HIV transmission mode – Baseline viral load – Baseline CD4 – Integrase baseline genotype 15

Acknowledgements Max-Planck-Institut für Informatik University of Cologne Thomas Lengauer Rolf Kaiser Nico Pfeifer Mark Oette Joachim Büch Saleta Sierra Aragon Prabhav Kalaghatgi Elena Knops Joachim Büch Maria Neumann-Fraune Eugen Schülter EuResist Eva Heger Francesca Incardona Claudia Müller Maurizzio Zazzi Nadine Lübcke Mattia Prosperi Institut für Immunologie und Genetik Kaiserslautern Medizinisches Labor Berg Martin Däumer Hauke Walter Alexander Thielen Martin Obermeier Berhard Thiele University of Düsseldorf Robert-Koch-Institut Björn Jensen Claudia Kücherer Alejandro Pironti