Dietary Interventions in Pollen-Related Food Allergy

In practice, it remains unclear what the best dietary approach is in subjects with pollen-related food allergy (PRFA). Our objective was to evaluate the effect of (1) dietary avoidance advice, (2) oral immunotherapy (OIT), (3) (heat) processing, and (4) consumption of hypoallergenic cultivars on frequency, severity, and eliciting dose of pollen-related food allergic reactions. A systematic search was conducted in PubMed, Embase, and Cochrane. All studies performing an in vivo investigation of one of the four interventions in adults with PRFA were included. Each study was assessed for quality and validity. Available data on frequency, severity, and eliciting dose of allergic reactions were extracted. Ten studies matched the eligibility criteria. No studies were retrieved on dietary avoidance advice. Two studies (N = 92) on apple OIT reported that tolerance was induced in 63% and 81% of subjects. Four studies (total N = 116) focused on heat processing. Heating was found to completely eradicate symptoms in 15–71% of hazelnut allergic and 46% of celery allergic individuals. Four studies (N = 60) comparing low to high allergenic apple cultivars revealed that Santana (and possibly Elise) apples seemed to cause milder reactions than Golden Delicious. In the awareness that overall level of evidence was low, we conclude that OIT, heat processing, and hypoallergenic cultivars may diminish or completely prevent allergic reactions in some but not all subjects with PRFA.

Primary dietary therapy for food allergy consists of the avoidance of triggering foods [5,6,10,11]. However, clinical guidelines on PRFA give no specific advice regarding avoidance of cross-reacting

Study Selection
After importation of all identified citations into EndNote and removal of duplicates, title and abstract screening, and subsequent full text screening were performed by two independent authors (E.A., A.M.v.D.). Selection was based on consensus; any discrepancies were resolved by consultation of other reviewers (S.A.L., H.v.O.-M., and T.-M.L.). In case of full text unavailability, we attempted to contact authors via email. References of selected articles, reviews and meta-analyses were hand searched and checked in the Scopus citation database for additional articles of interest.
All articles in English, Dutch, German, French, Spanish, and Italian were assessed. For inclusion, the study population had to meet three criteria: (1) ≥80% of the participants were 18 years or older; (2) subjects had a convincing history of hay fever or a positive SPT or ImmunoCAP to at least one type of pollen extract; and (3) subjects had a history of allergic reactions to foods known to cross-react with pollen as well as sensitisation (SPT or CAP) or positive challenge test to the food concerned. Studies were further assessed if they investigated at least one of the determinants of interest.
Studies evaluating immunotherapy other than OIT with eliciting food were excluded, as were studies where low allergenic cultivars were not compared to high allergenic cultivars. We also eliminated non-original studies (reviews, editorials, and expert opinions), conference abstracts, case studies, animal studies, post-mortem studies, etiologic, diagnostic, and prognostic studies, in vitro studies, and in vivo studies where allergy was only evaluated by SPT.

Data Collection
Two authors (A.M.v.D., S.A.L.) independently collected and recorded study characteristics on a predefined checklist, comprising the items author, setting, time frame, study design, study population, method of intervention, method of outcome measurement, and reported outcomes. In some studies, we evaluated part of the total study population because outcomes regarding our determinants of interest were only available for a subgroup of subjects.
For OIT, we obtained data regarding the frequency of achieved tolerance and tolerated dose at final follow-up. For processing and consumption of hypoallergenic cultivars, data on the number of subjects with no allergic reactions after intervention, on symptom severity, and on the eliciting dose were extracted. In order to improve comparability of results from individual studies, the proportion of subjects with an allergic reaction, the median VAS score for symptom severity, and the median dose eliciting symptoms were calculated from available data where possible.

Risk of Bias Assessment
The validity of included studies was assessed for the part of the study population considered relevant for our research question. The Robins-I tool [19] was used to evaluate seven potential sources of bias: bias due to confounding, bias in selection of participants into the study, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in measurement of outcomes, and bias in selection of the reported results. Two authors (A.M.v.D., S.A.L.) performed an independent evaluation and discussed disagreements to reach consensus. Each article received a final risk level of "low risk", "moderate risk", "high risk", "critical risk", or "no information".

Synthesis of Results
Because of evident heterogeneity in methodology and reporting between studies, it was considered inappropriate and infeasible to pool results. A qualitative synthesis of available results was therefore performed. No statistical analyses were conducted.
The overall level of the evidence per study outcome per intervention of interest was assessed using the GRADE system [20] and categorised as high quality, moderate quality, low quality or very low quality.

Study Selection
Our search yielded 6081 unique citations ( Figure 1). Screening of title, abstract, full-text, and related citations provided ten articles suited to address our research question, including one article found via reference checking.
All included subjects reported allergy to apple, hazelnut, celery, or carrot and had a history of pollen allergy or were sensitised to birch pollen (and additionally mugwort pollen in one study [23]).
No studies were obtained regarding the effect of dietary avoidance advice on frequency, severity, and eliciting dose of allergic reactions.
Two studies, including one randomized controlled trial (RCT), focused on OIT with increasing doses of Golden Delicious apple [21,22] in a total of 92 subjects. Both reported the number of subjects in the intervention group that achieved tolerance to apple and that could consume other Rosaceae fruits after a follow-up period of respectively 48 weeks [21] and 8 months [22]. One study provided information on median tolerated dose [22]. Neither study evaluated permanent tolerance, generally referred to as sustained unresponsiveness after a period of discontinuation of regular apple consumption [14]. No studies were found to evaluate OIT with other foods in our study population of interest.  In four studies with 116 subjects in total, authors reported on the effect of heat processing of hazelnut [25,26], celery [23,24], carrot [24] and apple [24]. In order to measure the effect, reactions in double-blind placebo-controlled food challenge (DBPCFC) with heated food was compared to reactions to raw food in DBPCFC or history. One study also investigated the effect of processing to celery spice on allergenicity of celery in this patient population [23]. The number of subjects with an allergic reaction to the processed food and their specific symptoms were reported in all studies, along with the information on the tolerated dose in three studies [23,25,26]. Other than heat processing and processing to celery spice, no other methods of processing appeared to have been evaluated in vivo by comparative raw versus processed food challenge.
Four studies compared the allergenicity of putatively high allergenic to putatively hypoallergenic apple cultivars, primarily assessing the difference in severity of allergic reactions by single-or double-blind food challenge in 60 subjects altogether [27][28][29][30]. Golden Delicious (GD), which was classified as the high allergenic cultivar in all studies, was compared to Santana apple in three studies [28][29][30], and Elise [30], Pink Lady [30], Topaz [29] and G-198/Orim [27] apples in one study each. All studies used various visual analogue scales to assess severity of reactions. Three studies provided information on the proportion of subjects who remained free of symptoms to the various apple cultivars [27,29,30]. The dose eliciting symptoms was discussed in only one study [28]. No studies were found to compare low to high allergenic cultivars for other foods than apple in subjects with PRFA.

Risk of Bias Assessment
The risk of bias in relation to our study question was moderate to high for all included studies, mainly due to possible confounding, selection bias (generally because only a subgroup of the total study population in some studies was relevant for this review), and bias in outcome measurement. Details of the assessment are presented in Table 2.

Synthesis of Results and Level of Evidence
A summary of our findings is found in Tables 3-5.

Oral Immunotherapy
After OIT with Golden Delicious apple, tolerance to apple was achieved in 63-81% of subjects (Table 3). In an RCT, Kopac et al. found the frequency of achieved tolerance after 8 months to be significantly higher in the intervention than in the control group (63% vs. 0%, p = 0.0001). In this study, authors also showed that the median tolerated dose was significantly higher at final follow-up compared to start of study in responders to OIT (N = 17, difference in median tolerated dose = 126 g, p = 0.0009), in contrast to the controls (N = 13, difference in median tolerated dose = 0 g) [22].
Tolerance to other cross-reactive fruits, vegetables, and nuts was reported to varying degrees (14-29%) in the OIT group by Kopac et al. [22]. Bouvier et al. stated that 98% of subjects who achieved tolerance to apple were able to eat other Rosaceae fruits (Table 3) [21].
Overall, OIT with apple appears effective in inducing tolerance to apple and some cross-reacting foods in individuals with PRFA. Level of evidence for these findings was very low according to GRADE-assessment (Table 6).

(Heat) Processing
In subjects with challenge-confirmed allergy to raw hazelnut, the percentage of subjects who were completely tolerant to roasted hazelnut varied from 15 to 71% amongst the two included studies (Table 4). Symptoms to both raw or roasted hazelnut were only mild, but the median dose required to elicit symptoms with roasted hazelnut appeared higher than with raw hazelnut [25,26].
For celery, Ballmer et al. found that 46% of subjects experienced no symptoms to cooked celery and no subjects (0%) tolerated celery spice. Six of 12 (50%) subjects had a moderate to severe reactions to raw celery, one of 11 (9.1%) to cooked celery and three of five (60%) to celery spice [23]. Only one case of celery allergy was examined by Bohle et al. and this subject had mild symptoms to raw and no symptoms to cooked celery [24]. There was insufficient information to compare dose thresholds between raw and processed celery.
Carrot was evaluated in three subjects and apple in one subject [24]. All subjects had mild symptoms to raw carrot or apple and no symptoms to cooked carrot or apple. No conclusions could be drawn regarding eliciting dose.
Overall, four studies on heat processing (mainly of celery and hazelnut) found that 15-100% of subjects with challenge-confirmed allergy to raw food experienced no symptoms to the same food when heated. GRADE-assessment resulted in a very low level of evidence for each of the evaluated outcomes (Table 6).

Hypoallergenic Cultivars
As described in Table 5, the percentage of subjects who remained completely asymptomatic after the final dose was described to be significantly higher for Santana apple than for Golden Delicious or Topaz apple by Kootstra et al. (54% vs. 7% vs. 7% respectively, p = 0.002) [29], but did not differ significantly between Santana, Golden Delicious, Elise, and Pink Lady apple according to Vlieg et al. [30], nor between Golden Delicious and G-198/Orim in Asero et al. [27].
All studies evaluating allergenicity of Santana apple showed that the symptom severity score after challenge with Santana apple was significantly lower than after challenge with Golden Delicious apple in subjects with pollen-related apple allergy (p < 0.05) [28][29][30]. Santana apple was also reported to be significantly less allergenic than Topaz apple in one study (p = 0.004) [29]. Vlieg et al., who compared severity of symptoms caused by Golden Delicious apple to those caused by Santana, Elise, and Pink Lady apples, conclude that Elise is also a low allergenic apple cultivar for subjects with PRFA [30]. On comparison of G-198/Orim to Golden Delicious apple in this patient population, both cultivars were found to cause the most severe reaction equally often [27].
Regarding the dose eliciting symptoms, Bolhaar et al. found that the quantities needed to provoke a reaction of equal severity were on average 30 times higher for Santana than for Golden Delicious apples (p < 0.001) [28]. Other studies did not report on this outcome [27,29,30].
Altogether, studies comparing low to high allergenic apple cultivars showed that Santana (and possibly Elise) apples seemed to cause milder allergic reactions than Golden Delicious apples in PRFA. The quality of evidence for the three investigated outcomes was graded as very low for the effect of this intervention.

Discussion
Overall, robust evidence regarding the effect of dietary interventions on the frequency, severity and eliciting dose of allergic reactions in subjects with PRFA is lacking. Evidence regarding the effect of specific dietary avoidance advice in this population is completely absent. Nonetheless, taking the low level of evidence into account, this systematic review of the available literature suggests that certain dietary treatments or adjustments can be beneficial for this group of patients. First of all, OIT with Golden Delicious apple seems to be effective in reducing the frequency of allergic reactions in subjects with birch pollen-related apple allergy, inducing tolerance in 63-81% of subjects. Secondly, heating of foods cross-reacting with birch or mugwort pollen appears to reduce allergenicity in subjects with PRFA, leading to complete prevention of allergic symptoms in 15-100%. Heating also possibly increases the dose threshold for symptom elicitation in pollen-related hazelnut allergic subjects. Finally, Santana and possibly Elise apples seem to cause less severe allergic reactions than Golden Delicious apples in subjects with birch pollen-related apple allergy.

Oral Immunotherapy
OIT with apple was found to result in tolerance in 63% [22] and 81% [21] of subjects, a varying but high response rate. Previous studies on the effect of OIT with plant-based foods were mainly focused on peanut [14,15] and were not included in this review because they were performed mainly in children and without the inclusion criterion of pollen allergy. However, these studies on peanut OIT showed that the rate of tolerance was found to range similarly to our review from 61-100% [14,15]. Sustained unresponsiveness in peanut studies, characterised by absence of symptoms to peanut despite irregular intake or prolonged avoidance, was achieved less frequently in 30-78% [14,15]. Although no evaluation of sustained unresponsiveness was performed for OIT with apple in the studies included in this review, Kopac et al. also suggest that tolerance may be transient, because no significant immunologic changes were observed and one subject experienced a relapse after discontinuing apple consumption during a holiday [22]. Therefore, OIT with apple in subjects with pollen-related apple allergy may be effective, but regular consumption after completion of the study is likely necessary to maintain tolerance.
Both Kopac et al. and Bouvier et al. reported that the majority of subjects with pollen-related apple allergy could consume other fruits and nuts after OIT with apple [21,22]. An explanation could be that these cross-reactive birch pollen-related foods share homologous aminoacid sequences, and therefore allergenic epitopes on the surface of these homologues [31]. Desensitisation to these epitopes in apple might result in desensitisation to these epitopes in other cross-reacting foods, inducing tolerance to more foods than just apple.

(Heat) Processing
The effect of heating on clinical presentation of PRFA has mainly been investigated for hazelnut and celery, and was found to eradicate symptoms in 15-71% of hazelnut and around 46% of celery allergic subjects. Furthermore, roasting of hazelnut resulted in higher dose thresholds [25,26] and boiling of celery caused fewer moderate to severe reactions [23]. Sensitisation to the birch pollenrelated PR-10 proteins, which are heat labile, explains the symptom diminishing effect of heat processing [32]. However, the effect of heating does not appear to be equivalent for all Bet v 1 homologues [24,32]. Where apple Mal d 1 undergoes a continuous unfolding process during thermal processing, carrot Dau c 1 and celery Api g1 do not begin to change structure until higher temperatures (respectively 28 • C, 43 • C, and 50 • C) [24]. Furthermore, Api g 1 returns to its native structure after recooling, where Mal d 1 and Dau c 1 do not [24]. Hazelnut Cor a 1 is reported to be heat-resistant below 100 • C [33]. These findings imply that, although heating may reduce symptoms in subjects with birch PRFA, this effect differs depending on the food. However, not only the level of heating appears to influence the allergenicity of pollen-related foods, as heating at the same temperature depleted allergenicity in some subjects but not in others. An explanation is that subjects may also be sensitised to heat-stable allergens, such as lipid transfer proteins (e.g., hazelnut Cor a 8, celery Api g 2) or seed storage proteins (e.g., hazelnut Cor a 9 and 14) [23,33,34]. (Co-) sensitisation to these allergens may explain why some subjects reacted to cooked celery [23] or roasted hazelnut [25,26]. In fact, Ballmer et al. provide support for this statement by demonstrating that not all celery allergic subjects were sensitised to the birch pollen-related Api g 1. Another explanation could be that the effect of heat processing foods in pollen-allergic subjects depends on the type of pollen sensitisation. For example, one study demonstrated that celery-mugwort sensitised subjects were IgE-sensitised to heated celery, whereas celery-birch sensitised subjects were not [35], indicating that mugwort-sensitised subjects may be more likely to react to heated celery.
Processing to spice does not appear to have the same effect as heat processing, but this was only studied for celery. All celery allergic subjects who underwent challenge with celery spice were found to be allergic to celery spice as well as to raw celery [23]. A previous in vitro study by Jankiewizc et al. also showed that it is possible to detect Api g 1, Api g 4 and celery CCD by specific IgE antibodies in celery spice [17], which supports the in vivo findings by Ballmer-Weber et al. Therefore, celery spice is not safe for pollen-related celery allergic subjects.

Consumption of Hypoallergenic Cultivars
To date, research with regard to the effect of consumption of alternative hypoallergenic cultivars on clinical symptoms in subjects with PRFA has focused on apple, showing that Santana apple appears to cause significantly less severe reactions than Golden Delicious apple [28][29][30]. These findings were later strengthened in a non-clinical setting, where around 40% of consumers with mild to moderate self-classified apple allergy reported having no symptoms to Santana apple [36]. Another apple which could be considered clinically preferable in this patient population based on the results of this review are Elise apples [30].
However, there were also some differences between similar apple cultivar comparisons in the different studies. For example, in Kootstra et al. subjects reached DBPCFC final dose (±100 g) of Santana apple significantly more often than of Golden Delicious apple [29], whereas no significant difference was found between the same cultivars in Vlieg et al. (final dose 120 g) [30]. Other factors to take into consideration which may influence severity of allergic reactions to apple are season [37], storage [28,30,38], consumption with or without peel [39], and intra-cultivar variation [27]. These elements were not part of this review, however.
It also becomes clear that classification of apple as hypoallergenic based on SPT [28,30] or Mal d 1 content [27] does not imply equally reduced symptomatology compared to high allergenic apples like Golden Delicious. Although Santana, Topaz, Pink Lady, and Elise were all classified as low allergenic [28,30], VAS scores of Santana were significantly lower than those of Topaz and VAS scores of Elise were significantly lower than those of Santana and Pink Lady [29,30]. Neither SPT nor Mal d 1 content seem to predict allergenicity of different apple cultivars as determined by food challenge [40].

Strengths and Limitations
In evaluating this review, the reader should remain aware of the low level of evidence due to small sample sizes, suboptimal study designs, and heterogeneity in intervention and outcome reporting between studies. The latter aspect discouraged pooling and meta-analysis. We would also like to point out that several studies had different primary aims than our review question. This meant that we had to focus on subgroups that dealt with our research question in some studies, possibly introducing selection bias as not all characteristics of the selected subgroups were available [23,26,27]. Finally, we were unable to find any studies in which the effect of dietary avoidance advice in practice on the frequency and severity of allergic reactions was evaluated.
Nonetheless, this is the first review analysing PRFA from a dietary point of view, in which we present an overview of potentially relevant dietary interventions to aid physicians, dietitians, and nutritionists in advising and treating these patients in practice. We feel our broad research question, extensive search strategy, transparent critical appraisal, and concise presentation of study characteristics and results will allow readers to make a conscious appreciation and interpretation of the available information.

Conclusions
In conclusion, subjects with pollen-related apple allergy may benefit from OIT with apple, which may additionally reduce symptoms to cross-reactive foods. Furthermore, apple allergic patients can expect less severe reactions if they consume the hypoallergenic apple cultivars Santana or Elise. Additionally, thermal processing of causative foods in subjects with PRFA likely reduces symptoms, but the effect size may depend on the food concerned. These findings can be used to advise subjects with PRFA on their diet, taking into account that the level of evidence is low.
In the knowledge that up to 90% of pollen-sensitised individuals suffer from PRFA [1][2][3][4], which can cause symptoms to a wide variety of fruits, nuts, and vegetables and thus deprive these individuals of valuable sources of vitamins, minerals, and fibre, more dietary intervention studies are necessary to consolidate our findings and evaluate the effect of avoidance versus allowance of causative foods, traces of causative foods and cross-reactive foods in the diet of patients with PRFA.