This Community is Archived.

While this community is no longer active, we invite you to review and recommend past posts and resources. Membership for this community is closed, but we hope you'll join us in one of the many other communities on GHDonline.

Moderators of Adherence & Retention and GHDonline staff

IAPAC conference highlight- using adherence to predict virologic failure

By Jessica Haberer, MD, MS | 20 Jun, 2012

Hi All,

I recently attended the IAPAC (International Association of Physicians in AIDS Care) Adherence Conference in Miami and one of the most interesting studies presented involved the use of adherence patterns to predict virologic failure. I've copied the abstract below and would interested to hear feedback from the community. One potential point of controversy is whether clinicians would be willing to use an algorithm to decide when to order a viral load. It could save dramatically on costs, but clinicians may worry about missing virologic rebound. Data on malaria diagnostics, for instance, has shown that clinicians still often treat for malaria even when the rapid tests (with high sensitivity and specificity) are negative.


Data-Adaptive Super Learning to Predict Viral Rebound based on Electronic Adherence Monitoring: An Analysis of the MACH-14 Cohort Consortium

Maya Petersen, Varada Sarovar, Anna Decker, Erin LeDell, Joshua Schwab, Robert Gross, Ira Wilson, Carol Golin, Nancy Reynolds, Robert Remien, Kathleen Goggin, , Jane Simoni, Marc Rosen, Mark van der Laan, Honghu Liu , David Bangsberg

Background. Electronic adherence monitoring has the potential to improve outcomes by triaging viral load tests and adherence interventions. Machine learning (automated algorithms for signal detection from complex data) may improve the accuracy with which viral failure can be identified.
Methods. We applied an ensemble machine-learning algorithm (“Super Learner”) to predict viral failure (rebound>400 copies/ml after suppression to <=400 copies/ml) using pooled data from the MACH14 consortium and compared the cross-validated accuracy of the resulting predictor to that achieved by traditional approaches. Medication event monitoring (MEMS) data were analyzed to predict viral rebound using: 1) average adherence; 2) logistic regression including average adherence and interruption >= 3 days; 3) Super Learner applied to 142 a priori selected candidate predictor variables, including basic clinical data and 134 adherence summaries (averages, nadirs of moving averages, variances, and frequencies and durations of interruptions). Super Learner employed internal cross-validation to data-adaptively select from among a user-specified library of algorithms including random forests, generalized additive models, Bayesian and Lasso regularized generalized linear models, and neural networks. Cross-validated area under the receiver operating characteristic curves [AUC] were calculated based on data not used in model fitting.
Results. 1137 patients with complete data for each predictor variable contributed 3149 HIV-RNA tests. 138 of 771 patients (18%) had at least one failure observed subsequent to 1810 suppressed HIV-RNA tests. The AUCs for simple average adherence and average adherence + 3 day interruption were 0.64 (95% CI 0.59-0.68) and 0.64 (95% CI 0.59-0.70), respectively. The cross-validated AUC for the Super Learner predictor was 0.72 (95% CI 0.68-0.76).
Conclusions. Super Learner analysis of electronic adherence data predicted viral failure with reasonable accuracy in a highly heterogeneous population of HIV infected individuals and could potentially be combined with real time monitoring to triage viral load testing and/or target patients for adherence interventions.



Ziad Khatib Replied at 7:24 AM, 4 Jul 2012

Thank you Dr. Haberer for posting the info of this very important study.
Petersen did conduct sophisticated type of analysis to show the efficiency of the super learner analysis of e-adherence in predicting VL failure.
Dr. Haberer addressed the point of the willingness of clinicians in using it. I can understand the concern of clinicians about missed viral rebounds, specially in limited-resource settings. Also it depends whether the clinics are using VL or only relying on CD4 cell count.

I would be curious to know the opinion of the community a propos the above discussion so far. Also if you have any more insights that you would like to share?


Jessica Haberer, MD, MS Replied at 10:41 PM, 7 Jul 2012

Thank you for your comment, Ziad. I'm posting the following comment on behalf of Maya Petersen:

Indeed, the results presented do not address the extent to which a targeted viral load testing strategy as compared to a strategy of joint viral load and CD4 testing of all patients at regular intervals, or a strategy CD4 only testing at regular intervals will result in delayed detection of virologic rebound, nor do the analyses address the clinical consequences of such a delay. We are currently conducting additional analyses aimed at achieving a better understanding of both the expected delay in detection of failure under a targeted testing strategy and its comparative costs.

Maxo Luma Replied at 8:44 PM, 12 Jul 2012

Ideally, to better monitor an HIV patient, jint CD4 count and viral load are needed. Since having a viral load that suppresses does not necessarily mean CD4 will go up. It depends on many other variables related to the patient. Just to support my statement, let me share with you the abstract of a paper that I just submitted on HIV, although not even accepted yet. Feel free to share your thoughts around that.

Attached resource:

Jessica Haberer, MD, MS Replied at 10:13 PM, 15 Jul 2012

Thanks for sharing your abstract, Maxo.

I agree that both CD4 and VL can be very useful for managing patient care; however, VL is quite expensive and often unavailable in resource-limited settings. Moreover, when it is available, it is often ordered per a preset schedule (e.g. every 6 months), not based on individual clinical needs. These issues drive the idea of using adherence patterns to predict who is likely to have virologic failure and thus need the actual test.

I am curious to hear from other members of the community as to whether they would be willing to use such a prediction algorithm.

This Community is Archived.

While this community is no longer active, we invite you to review and recommend past posts and resources. Membership for this community is closed, but we hope you'll join us in one of the many other communities on GHDonline.

Moderators of Adherence & Retention and GHDonline staff