An eight-gene blood expression profile predicts the response to infliximab in rheumatoid arthritis

Abstract

Background: TNF alpha blockade agents like infliximab are actually the treatment of choice for those rheumatoid arthritis (RA) patients who fail standard therapy. However, a considerable percentage of anti-TNF alpha treated patients do not show a significant clinical response. Given that new therapies for treatment of RA have been recently approved, there is a pressing need to find a system that reliably predicts treatment response. We hypothesized that the analysis of whole blood gene expression profiles of RA patients could be used to build a robust predictor to infliximab therapy.

Methods and findings: We performed microarray gene expression analysis on whole blood RNA samples from RA patients starting infliximab therapy (n = 44). The clinical response to infliximab was determined at week 14 using the EULAR criteria. Blood cell populations were determined using flow cytometry at baseline, week 2 and week 14 of treatment. Using complete cross-validation and repeated random sampling we identified a robust 8-gene predictor model (96.6% Leave One Out prediction accuracy, P = 0.0001). Applying this model to an independent validation set of RA patients, we estimated an 85.7% prediction accuracy (75-100%, 95% CI). In parallel, we also observed a significantly higher number of CD4+CD25+ cells (i.e. regulatory T cells) in the responder group compared to the non responder group at baseline (P = 0.0009).

Conclusions: The present 8-gene model obtained from whole blood expression efficiently predicts response to infliximab in RA patients. The application of the present system in the clinical setting could assist the clinician in the selection of the optimal treatment strategy in RA.

Conflict of interest statement

Competing Interests: Patent application pending.

Figures

Figure 1. Methodology for building and validating a robust microarray predictor.

The construction of robust microarray-based predictors must necessarily follow a series of steps in order to avoid analytical biases and ensure a real applicability of the model. First, the original sample is split in two subsamples: the training sample and the validation sample. In the training sample we seek to find the optimal classifier; complete cross-validation (leave-one out cross validation in our case) gives an unbiased measure of the power of each tested model. In the case that we find similarly good performing models, a resampling method (i.e. permutation testing) can be used to objectively select the most robust between them. Only once we have chosen the optimal model we will apply it to the validation sample. Since we have not used the information from this independent sample in building the predictor, the accuracy determined from this sample set is an optimal estimation of the power of the model in a real setting. A resampling method (i.e. bootstrap analysis) can be used to estimate the confidence intervals associated with the predictor accuracy.

Figure 2. Error rates associated to different parameter values for kNN classifier.

From left to right, predictor models with increasing number of genes and increasing number of nearest-neighbours are evaluated in the training dataset using LOOCV. The 8 gene model under 3, 4 or 5 nearest neighbours (green color) were found to be the optimal classifiers with only 1 patient misclassified out of 29 (0.034 error rate).

Figure 3. Percentage of inclusion of all genes selected through LOOCV.

The present plot shows the percentage that each gene is selected amongst the top 8 genes after 29 rounds of LOOCV. It can be seen that, from all genes, the 8-predictor gene group is systematically selected indicating a strong correlation with the outcome. The remaining genes seem to be selected on a random basis.

Figure 4. Flow cytometry CD4+CD25+ lymphocyte counts from responders and non responders to infliximab at weeks 0, 2 and 14.

Non responders had a significantly lower CD4+CD25+ lymphocyte fraction than responders at baseline (P = 0.0009). During the treatment this CD4+ subpopulation increased, ending with similar levels to the responder group.