10 participants with FPIES exhibiting reactions to more than 2 foods will be recruited.

Participants will have their blood drawn and be evaluated with a new blood assay that screens a large number of foods (more than 20) in a culture plate. Participants will be asked to eat the identified safe foods by the the blood assay.

A Receiver operating characteristic (ROC) curve was built to estimate the NPV. A random-effects logit model was used to model the binary outcome (safe or trigger food) as a function of the 9 biomarker measurements in the assay. (Expression Value = Relative fold change of 9-gene expression panel in response to food treatment, divided by fold change in response to LPS treatment, multiplied by 1000). The random effect in the logit model took into consideration the correlated data measured within the same subject. A cluster ROC curve analysis was used to assess the precision of the assay. Specifically, the area was computed under the cluster ROC curve (AUC). A threshold to obtain the NPV was selected based on inspection of the ROC curve.

At the initial visit, participants had their blood drawn, which was assayed. On average, participants came in up to 4 weeks later after the test results were ready. Participants were then asked to trial a new food each week for up to 7 weeks.

The same ROC curve described for the primary outcome was used to derive the PPV. At the initial visit, participants had their blood drawn, which was assayed. On average, participants came in up to 4 weeks later after the test results were ready. Participants were then asked to trial a new food each week for up to 7 weeks.