Predicting a durable immune response

Science Spotlight

Predicting a durable immune response

From HIV Vaccine Trails Network and Vaccine and Infectious Disease Division

March 20, 2017

Clinical trials for HIV vaccines aiming to assess the safety and immune response of volunteers to the candidate vaccine are currently underway. Evaluating the magnitude of that response over time is an important factor that should be considered in these trails. Data gathered from a trial could help the field decide whether to continue pursuing a certain vaccine candidate or to move on to new platforms. Looking at vaccine response at early time points (2 weeks after) or later (6-12 months after) gives insight to how long the protective effect of a vaccine may last. However, it is not always easy to follow all participants over the full time course needed to truly evaluate immune durability (6 months plus).  Therefore, a group from the division of Vaccine and Infectious Disease at Fred Hutch and researchers from University of Washington sought to determine whether including immune durability predictors in the data collected during HIV vaccine trials would predict outcome or need for further treatment.  They characterized cellular and humoral immune response durability using data from various clinical trials looking at validated biomarkers. The specific objectives of the study were to predict at a person’s sero-positivity and magnitude of immune response at 6 months based on data obtained from the two-week time point for the same biomarker.  

Graphs of correlations between biomarkers.

Graphical representation of observed and predicted immune responses from a biomarker in a HIV vaccine trail.

Figure provided by Dr. Huang.

The current study analyzed clinical trail results from 3 studies, covering 7 vaccines and 201 participants. From these studies, 10-27 cellular and humoral immune-biomarkers were measured. Data showed a wide range of responses over time for each biomarker. In general, antibody responses waned over time with moderate responses at week 26 and low to negative antibody detection at week 52. T cell responses appeared to be more heterogeneous at both time points. Spearman correlation coefficients of observed data at week 26 versus 52 of the same biomarker were positive, indicating a strong correlation between week 26 results compared to week 52. When separate biomarkers were compared to each other the results were less correlative than intra-marker comparisons. 

To model the week 52 responses, regularized random forests (RRFs) of 5000 classification or regression trees were used to model the predicted positivity or magnitude, respectively. For most biomarkers, the predicted response was reasonably accurate. As expected, biomarkers had heterogeneous prediction accuracy. The group also looked at increasing response predictability by adding baseline demographic variables or other week 26 immune responses as predictors, however this did not dramatically improve predictions.Like positivity, predicted immune response magnitude at week 52 varied from biomarker to biomarker. Again, adding other week 26 immune responses as predictors did not improve outcome for the majority of biomarkers. This analysis showed that the positivity and magnitude of the six month durable immune response was best predicted by the same biomarker’s two-week response level, however results varied across studies and biomarkers. This data, based on real data, supports the notion of using two-week data to predict durable immune response in HIV vaccine trials, especially where only partial participant durable response can be measured. This study is the first to systematically quantify the role of vaccine proximal immune response in predicting immune response six months later. Dr. Huang, the first author of the paper stated, “this study provides important evidence that the efficiency in the analysis of durable vaccine-induced immune responses (e.g., collected at 6 months after vaccination) could be increased (by up to 20-25%) via incorporating immune responses collected at 2-weeks post vaccination, especially when immune response durability is only assessed in a sub-sample of vaccine recipients”.

Funding for this study provided by the National Institutes of Health. 


Huang Y, Zhang L, Janes H, Frahm N, Isaacs A,Kim JH, Montefiori D, McElrath MJ, Tomaras GD, Gilbert PB. 2017. Predictors of durable immune responses six months after the last vaccination in preventive HIV vaccine trials. Vaccine, 35, 1184-1193.