In this study we investigated sIgE-tests and SPT to cereals in grass pollen allergic subjects clinically tolerating cereals. This approach is rare possibly because there has been a strong focus on developing sensitive tests for food allergy, whereas the diagnostic specificity has attracted less attention. Cross-reactivity between pollen and foods is important for some allergens, but for others it is frequently not correlated to clinical food allergy [10, 15–17].
Grass pollen allergy is one of the most frequent inhalant allergies with an estimated prevalence around 20% throughout the world, though it differs geographically [18, 19]. With an estimated prevalence of grass pollen allergy of 20% of the general population our data could suggest that 9% would be positive in SPT for a cereal and 4-7% in the tests for sIgE. If such test results are used without knowledge of the primary grass sensitization and the low specificity of the diagnostic tests, serious consequences could occur. One aspect is a gross overestimation of cereal allergy prevalence in epidemiological studies, when sensitization is used as a proxy for clinical allergy and proper follow-up is not possible. In several epidemiological studies, strikingly many IgE positives to wheat were in fact found . Another aspect relates to unwarranted dietary restrictions or needless DBPCFC in the daily practice. In the few studies investigating grass pollen allergy and cereal or peanut allergy, the food-allergy-status of the subjects has not been known with certainty [6, 11, 17, 21] or the test groups were small [8, 20].
In our study the positive reactions in both SPT and specific IgE tests all occurred in clinically tolerant subjects under the basic assumption that the result of OFC is a true representation of the clinical status. It might be criticized that we did not perform DBPCFC. Instead we used OFC, which we consider adequate since none of the subjects had any history of reacting to the foods tested, and the OFC is anyway recommended subsequent to a negative double-blinded challenge . The reason for using raw cereals was that uncertainty still remains as to whether the allergenicity is reduced during heating, as seen for fruits and vegetables cross-reacting with birch pollen, even though Pastorello et al.  found that cooking did not induce significant alterations of the protein structure of the true allergens. It could have been interesting to have included a third group of wheat allergic subjects for comparison but this was outside our aim and not easily done due to the low prevalence . Previous studies identified wheat allergens with molecular weight at 12-16 kDa [3, 24] and the major grass pollen allergens in group 1 and 5 range from 27-38 kDa [25, 26]. Despite this, common epitopes were identified in wheat and grass pollen as high molecular weight proteins [16, 17]. This suggests that the wheat proteins responsible for wheat food allergy are not the same as the wheat proteins responsible for cross-reactions in grass pollen allergics .
In immunological terms the positive reactions are likely to be specific since none of the 20 healthy controls had any positive reactions in any of the tests. In the SPT we used fresh-made suspensions of flour and saline which could be a source of error. Due to the limited characterization of wheat allergens, commercially available extracts are difficult to compare [3, 6], and freshly-made suspensions are used in the routine work at our Allergy Clinic. For the standardized inhalation allergen extracts used for SPT a cut off value of 7 mm2 is applied, but is this valid for non-commercial extracts? We would argue that without better characterization and the unsatisfactory diagnostic accuracies of SPT and sIgE-test [1, 6]; it is interesting to look at any reactions in the test, especially when all 20 healthy controls had negative SPTs-not a single reaction. The comparison of SPT and sIgE is hampered by the use of both non-commercial and commercial extracts. We used the cut off at 0.35 kUA/L, even though both Immulite and ImmunoCAP are able to measure IgE levels down to 0.1 kUA/L. Our data shows that a cut off lower than 0.35 kUA/L would not benefit the specificity. A cut off at 0.70 kUA/L (the inclusion criteria for grass pollen allergics) would improve the specificity, but the consequence for the sensitivity is unknown and was not studied in our design. The false positive rate-in clinical terms-was ranging from 0.20-0.46 of the three tests using the OFC results as the gold standard. The clinically irrelevant sensitization to cereals is often secondary to pollen sensitization showing in the level of positive reactions to grass pollen and wheat. We found that the mean SPT wheal size and the level of sIgE to cereals were much lower than to grass. Matricardi et al.  also found this significant difference in the level of sIgE. Furthermore they found that sensitization to wheat most frequently was secondary to pollen sensitization. They did not confirm the observed sensitization to foods by food challenge and therefore cannot conclude about the clinical relevance of these sensitizations. In 1995 Jones et al.  found that only 1 of 6 grass pollen allergic subjects tolerating cereals had SPT wheals to cereal grains > 3 mm, whereas all 6 had grass SPT wheals > 3 mm, supporting our assumption that the difference in level of sIgE or wheal size of SPT between the primary and secondary sensitization may indicate the clinical relevance. However, the difference in level of sIgE in general cannot predict the clinical relevance , as the cross-reaction to nuts in birch pollen allergic subjects is of high clinical relevance even though the level of sIgE to birch and nuts differ .
We saw a cluster of reactions to wheat, rye and barley more frequently than to oat, maize and rice. Wheat, rye and barley are taxonomically from the same tribe, where oat, maize and rice belong to three other tribes, with only oat belonging to the same subfamily as wheat, rye, barley and grass [8, 26]. This may explain why we see this cluster, which was also observed by Jones et al.  with 41-57% positive to wheat, rye and barley compared with 23-28% positive to oat, maize and rice. We cannot compare these results directly because some of the patients in Jones' study suffered from wheat allergy.
We also included peanut in spite of the absence of close botanical relation with cereals or grass pollen. Mortz et al.  found that 96% of Danish school children that were sensitized to peanut also had a concomitant reaction to grass pollen. In accordance with our findings they also found a lower sIgE level to peanut than to grass pollen, in grass pollen sensitized subjects with no clinically relevant peanut hyper-sensitization. Guilloux et al.  investigated peanut allergic patients some of which were also grass pollen allergic, and found a median sIgE level to peanut > 100 kUA/L (Immulite) and 49 kUA/L (ImmunoCAP) in the peanut allergic patients, whereas the median peanut sIgE was 0.11 kUA/L and 0.68 kUA/L respectively, in grass hay fever patients tolerating peanuts. The cross-reaction could also be due to Bet v 1 homologues in peanut, since all the patients positive to peanut were birch pollen sensitized, except one in the ImmunoCAP.
Reactions to CCD offer a good explanation on the observed cross-reactivity between grass pollen and cereals and peanut because of limited clinical relevance of CCD in the majority of grass-pollen-allergic patients [12, 29–31]. Our data with bromelain suggests that CCD might explain the reactions to cereals and peanut in only a few cases, however. It would be of interest to study this in detail by immunoblot assays and inhibition analyses. Constantin and co-workers performed a study of IgE-reactivitiy to components from grass pollen and wheat in different patient groups, of which was also included a grass pollen allergic, anamnestically food allergy-negative group of patients . Interestingly, 65% of these patients reacted to wheat flour in ImmunoCAP, but only 24% of the patients reacted to a wheat component (profilin), leaving the majority of the false-positive reactions to wheat unexplained. One could speculate that procalcin-another grass pollen pan-allergen-could also participate as a non-clinical relevant allergen .
It could be anticipated that the progress in wheat allergy diagnostics where single allergens responsible for wheat food allergy and baker's asthma are being characterized [1, 3, 4, 24, 32–34] may improve the specificity of such tests compared with whole extracts.
In conclusion, grass pollen allergic patients with proven food tolerance may express significant sensitization (albeit depending on the test system) to cereals and peanut. Overall 45%, 19% and 32% reacted to one or more cereals in SPT, Immulite and ImmunoCAP, with wheat, rye and barley causing the most frequent reactions. Likewise, peanut caused 9%, 13% and 23% false positive reactions. Negative food challenges for exclusion of clinical allergy are occasionally required in routine clinical work and clearly in epidemiological studies.