Adverse events associated with peanut oral immunotherapy in children – a systematic review and meta-analysis

While peanut oral immunotherapy (POIT) represents a promising treatment for peanut allergies in children, safety concerns remain a common barrier to widespread adoption. We aimed to systematically assess available evidence to determine the risk and frequency of adverse events occurring during POIT, and examine study-level characteristics associated with their occurrence and severity. A systematic search of MEDLINE, EMBASE, and Web of Science was conducted through April 2019. Controlled and non-controlled studies evaluating POIT were eligible. Twenty-seven studies, involving 1488 subjects, were included. Adverse events to POIT were common and led to treatment discontinuation in 6.6% of children (95% CI 4.4–9.0; 27 studies, I2 = 48.7%). Adverse events requiring treatment with epinephrine occurred among 7.6% (4.5–11.4; 26 studies, I2 = 75.5%) of participants, at a rate of 2.0 per 10,000 doses (0.8–3.7; 15 studies, I2 = 64.4). Use of a rush treatment phase and targeting a higher maintenance dose were associated with a higher risk and frequency of epinephrine use, while using co-treatments in addition to POIT was associated with a lower risk of treatment discontinuation due to adverse events. While adverse events to POIT are common, this study provides promising explorative evidence that certain modifications to existing treatment protocols could significantly improve treatment outcomes.

Risk of adverse events resulting in treatment discontinuation. The overall risk of treatment discontinuation due to adverse events was 6.6% (95% CI 4.4-9.0%; I 2 = 48.7%; N = 27 studies, Fig. 1), with no evidence of publication bias relating to small study effects (p = 0.352, S2 Figure). Risk of treatment discontinuation due to adverse events was lower among studies which included co-treatment alongside POIT (1.4%; 0-5.2%) compared with studies which used POIT alone (8.5%; 6.5-10.8%; p = 0.003) (Table 2, S3 Figure). When stratified according to co-treatment type, a lower risk of treatment discontinuation due to adverse events was noted among studies that administered probiotics or antihistamines, but not omalizumab ( www.nature.com/scientificreports www.nature.com/scientificreports/ meta-regression, a higher risk of treatment discontinuation due to adverse events was associated with higher baseline peanut specific psIgE (p = 0.0059), and a higher percentage of children with asthma (p = 0.0402) (S5 Table, S4 Figure).
Adverse events requiring treatment with epinephrine. The overall risk of an adverse event requiring treatment with epinephrine was 7.6% (4.5-11.4 I 2 = 75.5; N = 26 studies, Fig. 3), while the overall frequency of adverse events requiring treatment with epinephrine was 0.20 per 10,000 doses (0.08-0.37; I 2 = 64.4%; N = 15 studies). In subgroup analyses, risk of an adverse event requiring treatment with epinephrine was higher among studies which included a rush phase (11.6%; 8.1-15.6%) compared with studies which commenced with a slow build-up phase (2.3%; 0.1-6.1%; p = 0.001) and was higher among studies employing a target maintenance dose ≥1000 mg (13.7%; 9.6-18.3%) compared with <1000 mg (4.0%; 1.1-8.2%; p = 0.001) ( Table 3, S11 Figure). Similar findings were evident regarding frequency of adverse events requiring treatment with epinephrine (S6 Table, S12 Figure). Following univariate meta-regression, an increased risk of adverse events requiring treatment with epinephrine was associated with increasing median baseline peanut specific IgE (p = 0.0247) (S7 Table, S13 Figure), while a higher frequency of adverse events requiring treatment with epinephrine was associated with a higher target maintenance dose and higher baseline SPT (p = 0.0243) (S8 Table, S14 Figure).
Likelihood of reaching target maintenance dose. The overall likelihood of reaching target maintenance dose was 80.9% (95% CI 74.2-86.8%; I 2 = 86.2%; N = 26 studies, Fig. 4). In subgroup analyses, the proportion reaching target maintenance dose was higher among studies which included co-treatment alongside POIT (95.0%; 87.6-99.6%) compared with studies which used POIT alone (72.7%; 64.7-80.0%; p = < 0.001) (S9 Table, S15 Figure). When stratified according to co-treatment type, an increased likelihood of reaching the target maintenance dose was noted irrespective of the type of co-treatment administered (S9 Table, S15 Figure). Following univariate meta-regression, a low likelihood of reaching the target maintenance dose was associated with a higher target maintenance dose (p = 0.0088) (S10 Table, S16 Figure). www.nature.com/scientificreports www.nature.com/scientificreports/ Likelihood of passing supervised exit ofc. The overall likelihood of reaching the end of the study and passing a supervised exit oral food challenge was 68.9% (63.5-74.1%; I 2 = 53.0%; N = 17 studies, Fig. 5). In subgroup analyses, the likelihood of passing a supervised food challenge was higher among studies which included co-treatment alongside POIT (78.7%; 68.8-87.3%) compared with studies which used POIT alone (65.6%; 58.5-72.3%; p = 0.035, S9 Table, S17 Figure). When stratified according to co-treatment type, an increased likelihood of passing a supervised food challenge was noted among studies that administered probiotics or omalizumab, but not antihistamines (S9 Table, S17 Figure). Following univariate meta-regression, a higher likelihood of passing a supervised food challenge was associated with higher baseline SPT (p = 0.0058) (S10 Table, S18 Figure). In a sensitivity per-protocol analysis, the overall likelihood of passing a supervised oral food challenge and the end of POIT treatment among those able to tolerate the maintenance dose was 88.8% (83.2-93.5%; I 2 = 68.2) (S19 Figure).

Discussion
This systematic review and meta-analysis of almost 1,500 peanut allergic patients who received POIT among 27 controlled and non-controlled studies quantifies the risk and frequency of serious adverse events associated with treatment. While POIT appears effective in achieving the immunological goal of desensitisation in large proportion of subjects, serious adverse events lead to treatment failure in 1 in 15 subjects. Further, potentially life-threatening reactions requiring treatment with epinephrine are experienced by approximately 1 in 13 subjects throughout all stages of treatment. While the risk of potentially life-threatening reactions appeared similar across treatment phases, the frequency of epinephrine use was low once subjects reached the long-term treatment phase. Most notably, we identified modifiable treatment protocol related factors such as the elimination of a rush phase, aiming for a lower target maintenance dose, or use of co-treatments in addition to POIT, that could substantially improve the safety and efficacy of treatment regimens and warrant evaluation in future clinical trials.
Our study findings significantly expand upon those of the recent meta-analysis published by Chu et al. 12 , through including an additional 15 studies and data on more than twice the number of participants receiving POIT. Despite differences in inclusion of controlled and non-controlled studies, Chu et al. observed an absolute risk of adverse events resulting in treatment discontinuation of 6.1% (2.9-12.9) compared to our observed risk of 6.6% (4.4-9.0%). Similarly, our observed risk of epinephrine use of 7.6% (4.5-11.4%) was similar to the risk of 8.2%    12 . This substantially differs from the reported prevalence of passing a supervised exit OFC following POIT of 56% (n = 320/574) 12 , based on the crude data included in their-meta-analysis. A second issue is their inclusion of an unpublished study (NCT01324401 [PNOIT]; 2018) that performed the exit OFC one month after treatment avoidance. This study provided the lowest risk estimate for passing a supervised exit OFC (RR 3.30; 0.60-18.23) and therefore lowered the overall risk estimate, but was not eligible for inclusion as it was evaluating sustained unresponsiveness. In contrast, our meta-analysis demonstrated a higher overall pooled likelihood of passing a supervised exit OFC of 69.1% (58.1-79.1%) among blinded RCTs and 68.9% (62.1-75.4%) among open-label studies. Determining the true expected treatment effect should be the focus of future studies as this is likely to be a key factor influencing decision making regarding POIT.
Despite investigating more than 10 study-level factors in sub-group analyses, Chu et al. identified that only the requirement for failing an entry OFC as an entry criterion was associated with an increased risk of anaphylaxis during POIT 12 . The inability to detect additional associations is likely the result of inadequate statistical power. This is in contrast to the novel findings of our meta-analysis where use of an initial rush phase was consistently associated with an increased risk of serious adverse events, while aiming for a higher target maintenance dose was associated with an increased risk for epinephrine use. Further, use of co-treatment in addition to POIT was associated with a reduction in adverse events resulting in treatment discontinuation.
Only three RCTs have been undertaken to examine the effects of different approaches to POIT. These include different target maintenance doses (300 mg or 3000 mg peanut protein daily) 42 , use of co-treatment (omalizumab vs. placebo) 30 , or changing psychological mindset regarding symptoms associated with POIT 27 . While these studies, which recruiting between 35 to 50 patients, provided some promising evidence that targeting a lower target maintenance dose, utilising a co-treatment, or providing a psychological intervention may reduce the risk and frequency of adverse events, they were insufficiently powered to detect any clinically significant differences. Notably, we were able to overcome this limitation of relying on single studies by pooling data across studies in the largest meta-analysis undertaken on POIT to date.
Few previous studies have attempted to evaluate patient-level factors associated with the risk of adverse events during POIT, including three trials using individual patient data 37,43,46 and one meta-analysis 12 . Two studies examined differences in patient characteristics according to frequency of overall adverse events or epinephrine use 43,46 . Findings have been largely inconsistent, with one study demonstrating that rhinitis, asthma, and baseline SPT were all associated with an increased risk of adverse events 46 , while the other demonstrated that higher baseline psIgE was associated with an increased risk of epinephrine use 43 . In the meta-analysis by Chu et al., only increasing age was associated with an increased risk of serious adverse events 12 . We observed an increased risk of  www.nature.com/scientificreports www.nature.com/scientificreports/ adverse events among studies with a higher proportion of participants with rhinitis, while a higher proportion of participants with eczema and higher median age were associated with increased risk of adverse events requiring treatment. Further, our meta-analysis provides evidence suggesting that both higher baseline psIgE and SPT may increase the risk of adverse events during POIT, but each of these findings requires clarification in further studies.
Strengths of our meta-analysis include the thorough literature search, inclusion of data from controlled and non-controlled studies, and use of appropriate statistical methods for pooling prevalence data and undertaking meta-regression analyses. Our study also has some limitations. Because this is a study-level analysis, it is not possible to make inferences regarding which individual patients are at higher risk of adverse events related to POIT. Confidence intervals for pooling risk estimates include both between-study and within-study variations, and should be interpreted cautiously. Future studies pooling individual patient data could overcome such limitations. A high level of heterogeneity was observed across some study outcomes, which we attempted to take into account using random effects models, and explored through performing subgroup comparisons. An additional limitation was the incomplete reporting of data on some outcomes or study-level characteristics across studies. For example, less than half of included studies separated adverse event data according to treatment phase. Further, definitions of adverse events varied across studies. Some studies reported adverse events regardless of severity, while others restricted reporting to those considered to be moderate or severe. Greater efforts should be made to standardise reporting of adverse events in future clinical trials to facilitate pooling of data. Lastly, in some situations the study protocol was modified part way during the study 31,35,43 , with outcomes data reported for participants overall, rather than separated according to use of different treatment protocols.   www.nature.com/scientificreports www.nature.com/scientificreports/ While the risk of severe adverse reactions requiring treatment with epinephrine might appear high at 7.6%, this must be balanced against the benefits of POIT in providing protection against accidental peanut ingestion. A longitudinal study of a population of children who had developed peanut allergy before the age of 4 years revealed that, of the children with initial non-life-threatening reactions, 44% had at least one potentially life-threatening reaction during follow-up 47 . Based on this meta-analysis, once children reached the long-term maintenance phase of POIT, the risk of experiencing a significant adverse event requiring treatment with epinephrine was 3.2%, while the frequency was extremely low at 9 episodes per 100,000 doses.
Our findings are novel and identify modifiable study protocol related factors that can guide development of future treatment protocols that are safer and more effective. Given the current evidence that OIT is superior to placebo in achieving the immunological goal of desensitisation, future clinical trials should be focused on improving treatment protocols with the aim of improving safety, while maintaining high levels of efficacy. While expense and administration-related challenges associated with the use of omalizumab as a co-treatment may prohibit widespread adoption in clinical practice, we observed evidence that lower cost co-treatments such as use of oral antihistamines or probiotics were also associated with improved treatment-related outcomes. The specific value of these co-treatments should be evaluated in appropriate clinical trials. Further, in situations where POIT is currently offered to patients, our study provides significant information regarding not only the risk, but also frequency, of adverse events and related outcomes.
Lastly, an interesting observation relating to various treatment protocols was the inclusion of various recommendations made regarding dosing restrictions aiming to minimise the risk of potential adverse events. For example, some studies recommend avoiding exercise within 3 hours of dosing 36 , avoiding a hot shower or bath  Percentage (%) Figure 5. Proportion of participants able to complete peanut oral immunotherapy and pass supervised exit oral food challenge.