A systemic search strategy (Supplementary Material S1) was developed by combining synonyms for the patient population and intervention using both keywords and medical subject headings. The patient population consisted of those with birch pollen allergy and birch pollen–related food allergy; subcutaneous or sublingual birch pollen–specific immunotherapy was used as an intervention. Our search was performed in the PubMed, Embase, and Cochrane libraries on 3 November 2022.

Citations from the PubMed, Embase, and the Cochrane libraries were imported into the Rayyan tool for removing duplicates and for screening. Two authors (JL and EK) independently screened the titles and abstracts. When a paper was deemed possibly relevant, the full text was also independently screened by these two authors. Selection was based on consensus, and discrepancies were resolved by two other authors (TL and PW). English language articles that met the following criteria were included: (1) subjects with a birch pollen allergy, (2) subjects with BPFA for at least one food, (3) those with either birch pollen–specific SCIT or birch pollen–specific SLIT as an intervention, and (4) studies in which the effectiveness of this treatment was evaluated in terms of food challenge. Studies focusing on food allergy immunotherapy, non-original studies (editorials and expert opinions), conference abstracts, case studies, and animal studies were excluded. Reviews were also excluded, but they were used to obtain additional articles of interest based on reference checking.

Two authors (JL and EK) independently recorded the characteristics of the selected studies using a predefined checklist, comprising the following items: (1) study information (first author, year of publication, and country in which the study was performed); (2) study design [randomized controlled trial (RCT) or comparative/single-arm prospective cohort]; (3) type of food challenge [double-blind, placebo-controlled food challenge (DBPCFC) or open food challenge (OFC)]; (4) type of food; (5) treatment group characteristics (number of patients and type of immunotherapy), if applicable; (6) control group characteristics (number of patients and type of control); (7) timepoint when the outcome was measured; and (8) type of reported outcomes. The extracted outcome measurements were changes in (1) severity of symptoms during challenge, (2) eliciting dose, and (3) food allergy–related quality of life. Improvement in the eliciting dose was defined as the percentage of patients who could tolerate at least one higher dose without symptoms during the last-performed food challenge compared with baseline.

The validity of the selected studies was assessed using the revised Cochrane risk of bias tool (RoB2) (10), which evaluated five domains of bias: D1, the randomization process; D2, deviations from intended interventions; D3, missing outcome data; D4, measurement of the outcome; and D5, selection of the reported result. The following information was assessed: D1, performance of randomization, observed baseline differences in patient characteristics, and concealment of allocation sequence; D2, awareness of the assigned intervention, deviations from the intended intervention due to the trial context, and whether the analysis used to estimate the effect of assignment to the intervention was appropriate; D3, availability of outcome data; D4, whether the method of measuring the outcome was appropriate, comparable between intervention groups, and insensitive to awareness of the received intervention; and D5, whether data analysis was prespecified. Each of the questions in the domains could be answered with “yes,” “probably yes,” “probably no,” “no,” and “no information, which led to a risk of bias per domain classified as ‘low’, ‘some concerns’, or ‘high’”. Furthermore, the overall risk of bias was determined. Single-arm studies scored high for domains 1 and 2 because there was no control group/randomization (treatment effect estimates concerned pre/post-treatment differences) and patients and caregivers were aware of the received intervention.

Due to evident heterogeneity between the studies in terms of design, timepoint when outcome was measured, type of food, type of immunotherapy, type of control group, and availability and measurement of the five outcomes, it was considered inappropriate to pool the results. Therefore, a qualitative synthesis of the available results was performed without producing a formal statistical summary. Studies with the lowest risk of bias with a control group were considered the most important, while studies with the highest risk of bias without a control group were considered only supportive. Furthermore, a distinction was made between the direction of the effect (positive, no effect, and negative) and the size of the effect in case of an effect (large or small). Based on clinical interpretation, the size of the effect was considered large when there was an improvement of at least 20%. In the summary of the effect of birch pollen AIT on BPFA, only objective results are shown when a study reported on both subjective and objective results, because these are more reliable.

Our search yielded 3,652 unique articles (Figure 1). After screening the articles by title, abstract, full text, and reference checking, 10 articles were included.

Details of the 10 selected studies can be found in Table 1. All the studies were conducted in Europe. In total, there were five RCTs (of which one was a sub-study of an RCT) and five prospective cohort studies. A control group was used in all RCTs and in two prospective cohorts. In four out of five RCTs, the control group was a placebo group. In the other studies, the control group consisted of patients without AIT. In the studies without a control group, a pre/post-AIT comparison of the outcome(s) was made.

Altogether, 475 patients were analyzed, of whom 320 received AIT and 152 did not. Of the 320 patients who received AIT, 127 served as their own controls. Six studies focused on SCIT, three on SLIT, and one on either SCIT or SLIT. The last study was a three-arm study, in which patients receiving SCIT or SLIT or a placebo were compared. An OFC was performed in six studies, a DBPCFC in three studies, and both OFC and DBPCFC in one study.

During treatment, outcomes were reported at timepoints between 6.5 and 48 months after the start of treatment, but in 6 out of 10 studies, they were reported at 12 months.

The overall risk of bias was moderate in three studies and high in seven. Therefore, we focused on the three studies with the lowest risk of bias; the remaining studies were considered to be supportive. The overall risk of bias was mainly high because of issues with domain 2, “Deviations from the intended interventions,” which were attributed to no correction of prognostic factors in non-randomized studies. Details of the assessment per outcome reported are presented in Table 2.

A summary of the study results regarding the outcomes of severity of symptoms during challenge and ED is provided in Tables 3 and 4. In both tables, a distinction is made between studies with the lowest and the highest risk of bias and those with or without a control group. In studies with a control group, treatment effects pertained to the comparison between treatment and control groups, and in studies without a control group, treatment effects related to changes from baseline. Henceforth, for the three studies with the lowest risk of bias, only the comparison between the treatment and the control groups is discussed subsequently. Within these studies, a total of 61 patients received AIT and 37 did not.

Birch pollen AIT seems to have had a positive effect on severity of symptoms during challenge in the three studies with a moderate risk of bias; two showed a positive effect (11, 13) on severity of symptoms and one no effect (12) (Tables 3, 4). The first study distinguished between objective and subjective symptoms and showed a decrease in both objective and subjective symptoms in both the treatment and the control groups. Although there was no significant difference in the decrease, the patients in the treatment group tended to have a higher numerical decrease in objective symptoms than those in the control group (11). The second study with a positive effect showed a rate of reduction in symptoms of 69% in the treatment group vs. 0% in the control group (13).

The study with no effect only mentioned that there was no significant difference in symptoms during challenge; no numbers were mentioned (12).

The effect of birch pollen AIT on the eliciting dose is unclear. Of the three studies with a moderate risk of bias, one showed a positive effect (13), one no effect (11), and one a negative effect (12) on the ED (Tables 3, 4). The study with a positive effect showed a 24-fold increase in the eliciting dose in the treatment group vs. no change in the control group. Furthermore, 23% of the patients in the treatment group vs. 0% of the patients in the control group reached ingestion of the highest dose without symptoms during the last challenge (13). The study with a negative effect showed a baseline mean ED that did not change in the treatment group but increased from 0.10 to 1.00 g in the control group (12).

Further research is needed to assess the effect of birch pollen AIT on FA-QoL. Only one study with a moderate risk of bias investigated the effect of birch pollen AIT on FA-QoL (11). This study used the validated food allergy quality of life questionnaire—adult form (FAQLQ-AF) and showed that there was no significant difference between patients receiving SCIT or a placebo. No studies with a high risk of bias investigated the effect of birch pollen AIT on FA-QoL

Supporting studies showed mostly positive effects. The seven studies with a high risk of bias were considered supportive, of which four (14–17) included a control group and three (18–20) did not (Table 4). When there was a positive effect on severity of symptoms during challenge, there was also a positive effect seen on the ED. Of the studies with a control group, three (15–17) showed a positive effect on severity of symptoms and the ED with mostly a large effect size and one (14) no effect on severity of symptoms during challenge and a negative effect with a small effect size on the eliciting dose. Of the studies without a control group, two (18, 19) showed a positive effect with a large effect size and one (20) no effect.

Overall, these high risk of bias studies supported the moderately positive effect of birch pollen AIT on severity of symptoms during challenge and indicated that the effect on the eliciting dose was more likely to be positive.

As mentioned previously, 7 (14–20) of the 10 studies had a high risk of bias, mostly due to domain 2, “Deviations from the intended interventions,” followed by domain 4, “Measurement of the outcome.” The high risk of bias in domain 2 was due to studies without a control group and studies with a control group but not adjusted for prognostic factors when no randomization was performed. The high risk of bias in domain 4 was attributed to the fact that, among others, studies performed an OFC only at the end of the study and not at the start (15, 19) or used a method in the treatment group that was different from that in the control group (16). Because most studies had a high risk of bias, the quality of evidence was low. To obtain the best possible estimation of effectiveness, we decided to use the three studies with the lowest risk of bias for the assessment and the studies with a high risk of bias only as supporting evidence.

Birch pollen AIT seems to have a positive effect on BPFA as evidenced by an alleviation of symptoms during challenge, and it remains unclear whether there is also a positive effect on the ED. That some studies found no or even a negative effect could be attributed to the fact that the included patients might not have had a pure BPFA but also a primary food allergy that did not reduce or even worsen during treatment and/or to the fact that there was an imbalance in the groups in this respect. Nowadays, by measuring both the PR-10 and the non-PR-10 components, it has become possible to differentiate between a pure BPFA and a primary food allergy (21). The study that showed that there was no effect on symptoms during challenge and a negative effect on the eliciting dose measured only Cor a 1 and Cor a 8 but not Cor a 9 and 14 (12). Therefore, it was unclear whether only patients with pure BPFA were included.

Another reason could be that the follow-up period was too short. In general, the effect of immunotherapy increases with a longer duration (22, 23). All three studies with the lowest risk of bias reported their results only after 1 year of AIT use.

Patient-reported outcome measures (PROMs) are often the best way of measuring patient symptoms and quality of life and can help reduce observer bias (24). Unfortunately, only one study investigated the effect of birch pollen AIT on the patient-reported outcome “FA-QoL” using the validated FAQLQ-AF questionnaire and showed that there was no significant difference between patients receiving SCIT or placebo at endpoint (11). As the primary burden on patients living with food allergy is a reduced QoL, treatment success in trials should also be defined by an improved QoL (25, 26). Future studies, investigating the effect of birch pollen AIT on BPFA, should therefore include QoL as an outcome.

In total, six studies investigated the effect of birch pollen SCIT (11–13, 16–18), three that of birch pollen SLIT (15, 19, 20), and one that of both birch pollen SCIT and SLIT (14) on BPFA. Because of the small number of studies and the high risk of bias, there is too little evidence to draw a conclusion.

In our systematic review, we found that almost all studies investigated the effect of AIT with either hazelnut or apple as a type of food. This is not surprising, as hazelnut and apple allergies are among the top three birch pollen–related food allergies reported in birch pollen–endemic areas (27). However, patients with BPFA are mostly allergic to multiple types of fruits, vegetables, and nuts. The effect on other foods remains unknown. Therefore, it is important to evaluate the effect of AIT on a broader range of BPFA foods.

We expected a better therapeutic effect of birch pollen AIT on foods with PR-10 components that are more homologous to Bet v 1. However, although the PR-10 components of apple and hazelnut are more homologous to Bet v 1 than those of soy (28), this review showed a better effect of birch pollen AIT on soy (11) than on hazelnut allergy (12). To confirm this hypothesis, more studies are needed that compare the effect of birch pollen AIT on multiple foods related to BPFA.

It is unknown how long the effects last after one discontinues AIT. Most of the included studies measured the outcome after 12 months of the start of birch pollen AIT, but none showed results after discontinuation of AIT. In 2003, Asero conducted a prospective cohort study to evaluate the long-term effect of birch pollen–specific SCIT on apple allergy after treatment cessation (29). In this study, 21 BPFA patients who discontinued birch pollen SCIT could tolerate apple. However, the effect appeared to decrease over time, since after 30 months of discontinuation, only 52% of the patients remained symptom-free. Further studies are needed to investigate the long-term effects of birch pollen AIT on BPFA so that clinicians can advise patients appropriately.

Apart from the effect of birch pollen AIT on BPFA, several studies showed the effect of other immunotherapies on plant food allergy. Two studies reported about an effective treatment of birch pollen–related apple allergy. Kinacyian et al. (30) showed that patients receiving SLIT with rMal d 1 required a significantly higher dose of rMal d 1 to induce OAS compared with the group that received rBet v 1 and placebo (p = 0.001), and Kopac et al. (31) showed that apple consumption induced a transient tolerance. Furthermore, studies from Japan (32) and Italy (33) reported the positive effects of Japanese cedar pollen–based SCIT and grass pollen SLIT, respectively, on plant food allergy. All of the above studies showed promising approaches for the effective treatment of plant food allergy, but these results should be confirmed before they are used in clinical practice.

Due to the high risk of bias and the heterogeneity of the included studies in terms of, among other elements, the study design, type of food, type and dose of immunotherapy, and method of assessing response to treatment, it was not possible to pool the results. Furthermore, symptoms were often not specified, and none of the studies reported the minimal clinically important difference (MCID) to define improvement, which made it difficult to interpret whether the differences found were clinically relevant (34). To provide the most reliable results, we focused on the three studies with the lowest risk of bias. Because of this selection, the total number of patients was small, with the total number of patients receiving AIT being 61 and those not receiving AIT being 37. However, the strengths of this review included its comprehensive search and methodological rigor, which also took into account patient-reported outcomes. This factor made this review the first systematic one to show the effect of BPFA on clinical and patient-reported outcomes.