Interpretation of ELISA Results
A priori, we determined that serum banked before the release of StarLink™ corn (GROUP C) comprised the negative controls for comparison with the cases (GROUP A). Serum from multiple allergic, or atopic, individuals (GROUP B) were tested to ensure that the presence of high IgE did not give a false-positive test result.
Figure 2 [opens in new window] shows the distribution of absorbance values that were documented in each group during the first FDA run. Duplicate samples were included for internal quality control and account for additional data points on the plots. There is expected variation around the run-specific diluent blank. For the 16 cases in GROUP A, the absorbance readings are between 0.06 to 0.11. For the five atopic samples in GROUP B, the absorbance readings are between 0.06 to 0.09, and for the 21 pre-1996 samples in GROUP C, the absorbance readings are between 0.08 to 0.12. Figure 3 [opens in new window] shows the same information for the second FDA run. The first run had a slightly higher reagent blank absorbance reading (0.07-0.10) than in the second run (0.05-0.06). The first run also showed greater overall variance in the absorbances in cases and a priori controls than did the second run. In both runs all cases have lower absorbance readings than the pre-1996 controls (GROUP C). Figure 4 [opens in new window] displays the third FDA run, which includes an additional case sample that arrived later than the other samples as well as several duplicates of previously run samples.
In all three runs, the readings from stored pre-96 controls are generally higher than the readings from freshly drawn serum (cases and atopic controls). It is not uncommon to see higher background absorbance readings in serum samples frozen and stored for longer periods (GROUP C or pre-1996 controls) than in fresher serum samples (GROUP B or atopic controls; Oliver 2000). There are no other consistent trends among the replicated serum specimens. The absorbance readings for all groups appear to reflect variability in the background range.
Using standard protocol (Rose 1992), a positive (reactive) ELISA test was defined by an absorbance reading that exceeds the cut-off value computed by multiplying the run-specific mean of GROUP C (determined a priori to be the negative control) by 2.5, as shown below.
First run 2.5 (0.098) = 0.245
A reactive ELISA test must exceed the absorbance reading of 0.245 for the first run.
Second run 2.5 (0.078) = 0.195
A reactive ELISA test must exceed the absorbance reading of 0.195 for the second run.
Third run 2.5 (0.171) = 0.428
A reactive ELISA test must exceed the absorbance reading of 0.428 for the third run.
The highest absorbance reading for the cases (GROUP A) was 0.107 for the first run, 0.081 for the second run, and 0.136 for the third run—each of which are considerably lower than their respective cut-off values calculated from GROUP C (i.e., 0.245, 0.195, and 0.428). Even if we recalculate the cut-off values using the atopic (GROUP B) mean values as our control reference, none of the case serum specimens exceed the cut-off values computed from the first run (0.198), the second run (0.168), or the third run (0.305). The reactivity of all serum samples with Cry9c or with control allergens (cat, grass, peanut) was also assessed by direct comparison with the reagent blank, which contained no serum at all. Any serum that produced an absorbance (optical density) reading less than twice the average reading of the reagent blank was considered "non-reactive", any serum that produced an absorbance reading greater than ten times the average reading of the reagent blank was considered "strongly reactive," and any serum that produced an absorbance reading between those ranges was considered "reactive." These results are summarized in Table 1 [opens in new window].
Figure 5 [opens in new window] demonstrates that the positive signal obtained with the Cry9c-immunized goat serum is three to four orders of magnitude greater than any of the human serum samples (GROUPS A, B, and C). This suggests that the ELISA was sufficiently sensitive to detect low concentrations of Cry9c antibody.
Figure 5 also shows the results that FDA reported of other serum samples from people with known allergies (i.e., cat, grass and peanut) that were analyzed at the same time as the CDC samples in this study. Using the same ELISA with cat, grass, and peanut allergens, FDA was able to accurately detect antibodies to the known substances causing allergies in these people. This finding serves as an additional internal quality control procedure.
We found very similar patterns when we reviewed the results that a University of Maryland laboratory obtained when they analyzed the same set of samples that CDC sent to FDA.
Our analysis was designed only to detect IgE antibodies that reacted with Cry9c, and we did not have a positive human serum control that reacted with Cry9c. It is possible that other antibodies to Cry9c were present, or that IgE antibodies were present in such lows levels in serum samples from case subjects that the ELISA could not detect them. It is also possible for people to have food allergies without any detectable IgE to the allergen (Ogura 1993). However, the FDA ELISA method was capable of detecting IgE antibodies to other allergens (cat, grass, and peanut) in every control sample tested, and these results were replicated by an independent laboratory.