Basophil activation test has high reproducibility and is feasible in the clinical setting

Abstract Background The basophil activation test (BAT) has high accuracy to diagnose peanut allergy and can reduce the need for oral food challenges (OFC); however, so far it has not been incorporated in clinical practice. Methods We assessed the reproducibility of BAT within the same laboratory and between two different laboratories and the feasibility of using BAT in the clinical setting. Results One hundred and two children being assessed for peanut allergy were tested on BAT (72 allergic, 30 sensitized tolerant). There was little internal variation (coefficient of variation <15%) in the BAT and a very strong correlation (R s > .95) between BAT performed across laboratories. The 2 BAT methods were strongly correlated but not interchangeable. In the cases of discrepancy, our in house BAT method was 100% accurate. BAT was feasible and well‐accepted by clinicians: no patient with positive BAT was referred for OFC, leading to reduction in the number of OFC required. Twenty one percent of patients who underwent OFC reacted to peanut. A negative BAT also encouraged the performance of OFC in sensitized children who would otherwise be considered allergic, 50% of whom did not react and incorporated peanut in the diet. Conclusions The BAT is a robust test that can reliably be transferred between laboratories; however, different BAT methods are not interchangeable. BAT was well integrated in the clinical decision‐making process in a specialized center.


| INTRODUC TI ON
Food allergy has become increasingly prevalent and severe in the recent decades, giving rise to increased awareness and increased need for testing. 1,2 In many cases, exposure to the allergen in a medically supervised and controlled environment in hospital during oral food challenge (OFC) is required to ascertain whether the child is allergic or not. However, OFC are resource-intensive and place the patient at risk of potentially severe allergic reactions and the need for OFC far exceeds current capacity of Allergy services.
We and others have previously demonstrated that the basophil activation test (BAT) has high specificity and sensitivity to diagnose food allergy. [3][4][5][6][7][8][9] For instance, for peanut allergy, BAT was accurate in 97% of cases and reduced the need for OFC in approximately 67%. 10 We confirmed the diagnostic performance of BAT to peanut in a large well-characterized cohort of children who participated in LEAP and associated studies. 11 Bringing BAT to clinic would enhance the accuracy and safety of food allergy diagnosis. However, BAT is still a research test not available to clinicians seeing patients with suspected food allergy in the majority of clinical settings. Different stages need to be achieved for the transition of BAT to the clinic, 3 including: 1. standardization of the methodology and reliability of its application in different laboratories; 2. technical validation and clinical validation of BAT; and 3. feasibility, for instance in terms of access to flow cytometry, transportation, and timely processing of samples.
In this study, we aimed to assess the consistency and reliability of BAT within the same laboratory and between two different laboratories and to assess the feasibility and acceptability of using BAT in the clinical setting.

| Study population
Two groups of subjects were tested in this study: adults with no known allergic diseases as healthy controls and children aged 6 months to 15 years being assessed for possible peanut allergy, that is, either had a history of reaction or unknown consumption of peanut and/or sensitization to peanut. Healthy adults were recruited following ethical approval (reference 14/LO/1699). Their samples were used for optimization and determination of intra-assay variability, all the other results were generated with samples collected from children with suspected peanut allergy who were recruited between 2019 and 2021 as part of the study "Diagnostic markers of clinical allergy versus sensitization to peanut" (10/H0802/044), as previously described. 10 The samples collected from healthy adult donors were stimulated with anti-IgE, fMLP, and buffer alone. The samples collected from children were stimulated with peanut extract in different concentrations, in additions to anti-IgE, fMLP, and buffer alone, as controls. Children underwent diagnostic assessment for peanut allergy, including clinical assessment, skin prick test (SPT), blood collection for specific IgE and BAT, and OFC if clinically indicated.
Ethics approval was obtained and informed consent from adults with parental responsibility and assent from children were obtained prior to any study procedures.

| Basophil activation test
The samples collected from healthy adult donors were used for technical validation of the assay and basophils were stimulated with anti-IgE, fMLP, and buffer alone. In the samples collected form children, basophils were stimulated with peanut extract in different concentrations, in additions to anti-IgE, fMLP, and buffer alone. For the BC-BAT, 50 μl of Dulbecco's PBS (pH 7.2) were initially added to reconstitute the dry-freezed antibodies (CD45-KO,   CD3-PC7, CRTH2 APC, CD203c PE, CD63 PB450 from Beckman Coulter) containing peanut extract (ALK-Abello) at serial 10-fold dilutions from 10 μg/ml to 0.1 ng/ml or anti-IgE or fMLP provided by BC, as controls. This was then followed by the addition of 50 μl of heparinized whole blood, and incubation at 37°C and 5% CO 2 for 20 min. An erythrolytic reagent, OptiLyse C (Beckman Coulter) was then added, and any excess and unbound antibodies were washed off with Dulbecco's PBS followed by a final centrifugation at 300 g for 5 min at RT. Flow cytometry was performed using CytoFLEX (Beckman Coulter), and data were analyzed using FlowJo software (version 10.6.2; Ashland). Figure S1 shows the gating strategies adopted for both BAT methods. Non-responders were defined by a %CD63+ Basophils following stimulation with anti-IgE of <5% and with fMLP of 5% or more.

| Standardization of flow cytometers
To reduce variability between the two Cytoflex platforms, we first optimized the gain settings of the cytometers. A gain titration was performed on the first Cytoflex to determine the optimal gain for each channel on the KCL Cytoflex using 8 peak rainbow calibration particles (P-RCP8-3.0; Kisker Biotech). This involved setting the FITC gain to 400 and adjusting all other fluorescence gains to 10 before recording 5000 bead events. Next, all fluorescence gains except FITC were increased to 20 and recorded as previously. This step was repeated by increasing the gain by 10 each time until reaching 100, then increasing the gain by 100 until reaching 1000, and finally increasing the gain by 250 until reaching the maximum gain of 3000.
Subsequently, the median and standard deviation of the 1st, 2nd, and 4th peak for each channel was extracted. The same method was then performed by setting the PE gain to 200, and adjusting all other fluorescence gains starting at 10 and repeating the entire procedure as described for the FITC gain titration.
For determination of minimum gains, the signal to noise ratio (S/N) defined as MFI (peak 2)/MFI (peak 1) was calculated and plotted as a function of gain. Additionally, the coefficient of variation (CV) of peak 4 was also plotted as a function of gain. To determine the minimum gain for each channel, the point at which the S/N and rCV plots plateau was selected as the minimum gain value. The final optimal gain value for each channel was then calculated as an average of the minimum gain values selected in the previous step.
Standardization was performed between the two Cytoflex platforms following the manufacturer's instructions (version B49006AP) using the optimal values selected above. Target median values were generated using Daily QC fluorospheres (B53230; Beckman Coulter) from a specific lot for each channel on the first Cytoflex. Resulting median fluorescence intensities were calculated from a total of six replicates and used as the standardization target value. The average MFI values were then matched as close as possible on the second Cytoflex using the same lot of QC fluorospheres to create a reference standardization file. Standardization was performed using the same lot of beads before each experiment on each Cytoflex.

| Statistical analyses
The distribution of variables was not normal as assessed by the Kolmogorov-Smirnov and Shapiro-Wilk tests, thus non-parametric tests were used. Mann-Whitney U test and Wilcoxon rank were used for comparison of the distribution of quantitative variables between independent and paired groups, respectively. Spearman correlation and Bland-Altman plots were used to assess the relationship between paired variables, namely the comparison of BAT results using two different methods tested across two different laboratories. We expected that the CV of the BAT was ≤15% and that the correlation of results obtained across methods and across laboratories was ≥0.90. ROC curve analyses were used to determined the diagnostic utility of the tests. The peanut concentration with the largest area under the ROC curve for each method was selected and the optimal cut-off defined by the Youden index was selected for the optimal concentration for each method. Most analyses were performed with SPSS 27.0 (IBM Inc) and graphs were designed using GraphPad Prism 9.0 (GraphPad Software, Inc).

| The basophil activation test show little internal variation
We assessed the variability and reproducibility of BAT by testing the same conditions repeatedly (n = 10) by two different operators. The coefficient of variation ± standard deviation was 3.48 ± 1.92% and 10.32 ± 3.69% for IH-BAT in the hands of a more and a less experienced operators, respectively ( Figure S2). We also compared the IH-BAT analyzed in two different cytometers (BD Canto II versus BC Cytoflex) in parallel and observed a high consistency in the results obtained, particularly if the cytometers were standardized before analyses (data not shown).

| The basophil activation test has high reproducibility when tested in separate laboratories
We recruited 102 children being assessed for possible peanut allergy (Table 1). We performed the IH-BAT in two different laboratories, one research laboratory, and one clinical diagnostic laboratory, using 65 samples from these children assessed for possible peanut allergy.
The IHBAT results obtained across laboratories were comparable (Table 2, Figure 1). The correlation of the results obtained between the two laboratories was very strong, with correlation coefficients above .95 for all allergen concentrations tested, and the Bland-Altman bias was very low. A second BAT method (BC-BAT) tested in parallel (n = 65) showed statistically significant differences between laboratories in some conditions, despite the strong correlation, both using CD63 and CD203c activation markers (Table S1, Figure S2).

| The results of the basophil activation test using two different methods are not interchangeable
We compared two different BAT methods head-to-head (IH-BAT versus BC-BAT) using samples of children assessed for possible peanut allergy (n = 102). Although the proportion of CD63-positive basophils following peanut stimulation was not significantly different between the two methods, the MFI for the activation marker CD203c was lower using the BC-BAT compared with IH-BAT in all conditions tested (Table 3, Figures 2 and S3). The correlation between the basophil activation using CD63 measured with the two BAT methods was strong and the bias calculated with Bland-Altman was low; however, the dispersion of the results was substantial. The correlation of results obtained using the second activation marker CD203c was also strong but the levels of CD203c expression measured with the BC-BAT method were systematically lower than the levels measured with the IH-BAT method.
There did not seem to be a systematic error between the methods; however, at the individual patient level, a discordance between the results obtained with the two methods was evident. For instance, patients tested negative with one method while testing positive with the other method, which can have diagnostic implications. Positivity of BAT results was defined for each method based on the optimal cut-offs determined by the Youden index in ROC curve analyses (n = 95 for IH-BAT and n = 82 for BC-BAT, as participants with non-responding basophils were excluded, Figure S5).
Considering only the patients that underwent OFC (n = 32), there TA B L E 1 Characteristics of study population. Data are presented as number and percentage for qualitative variables or median and interquartile range for quantitative variables. were four patients (out of 32, 12.5%) for whom the two BAT methods provided opposite results ( Table 4)

| Integration of the basophil activation test in clinical decision-making
In the last stage of the project, the results of BAT were fedback to the referring clinician for 79 participants, who were referred by 21 F I G U R E 1 Head-to-head comparison between the in-house basophil activation test (IH-BAT) across laboratories, a research laboratory (KCL) and a diagnostic laboratory (DxLab) in terms of: basophil activation using CD63 (A) and CD203c (B) in a variety of stimulating conditions (RPMI alone, peanut extract 10 and 100 ng/ml, anti-IgE, and fMLP); correlation of CD63+ basophils (C) and stimulation index of CD203c (D) following stimulation with 100 ng/ml of peanut extract; Bland-Altman plot of difference over average of basophil activation following stimulation with 100ng/ml of peanut extract using CD63 (E) and CD203c (F).

| DISCUSS ION
The BAT to peanut has previously shown high accuracy (97%) in the diagnosis of peanut allergy and ability to reduce the need for OFCs by two thirds 10   Note: Patients who did not undergo oral food challenge or had nonresponding basophils were excluded.