Basophil activation test: How many basophils are enough?

Abstract

The basophil activation test (BAT) is gaining prominence as an in vitro diagnostic tool for IgE-mediated allergies.¹ For food allergies, the European Academy of Allergy & Clinical Immunology (EAACI) now recommends the use of BAT prior to oral food challenges.² EAACI recently published a consensus report to standardize BAT protocols for better inter-laboratory comparisons and quality assurance.³ Specifically, the report defines an activated basophil as one with CD63 expression above the 97.5th percentile of a resting basophil population (Figures 1A and Figure S1).³ A positive response is indicated when more than 5% CD63+ basophils are measured in a stimulated sample, although this cutoff may be adjusted for specific allergens, generally 5%–15%.³

However, a critical factor remains unaddressed: basophil counts. Most studies enforce a minimum of 200–1000 basophils, with 500 basophils broadly accepted as the lower limit (basophils represent <1% of leukocytes).⁴ Recent reports have challenged this consensus, suggesting that dose–response curves in peanut-allergic individuals can be assessed with as few as 50 basophils.⁵ A universal approach to BAT analysis that accounts for basophil count is therefore urgently needed to boost confidence in test results across assays and laboratories, particularly in patients with low basophil counts (e.g. basopenia) or for low-volume samples.

Herein, we generalize activation cutoffs for any given basophil count by considering their confidence intervals, arising from the limited number of observations (Figure 1B). This effectively introduces a ‘confidence margin’ around the cutoff, within which results are considered inconclusive (Figure 1C). A result is unequivocal only if the CD63+ fraction falls either above or below the confidence interval. For instance, in cases with low basophil counts, activation levels must be well above the cutoff to statistically conclude a positive result, owing to the stochasticity inherent to small populations (Tables S1 and S2). Conversely, above 500 basophils, the result interpretation is close to that of a conventional fixed cutoff.

To validate our proposed approach, we statistically analysed a BAT dataset comprising 299 whole blood samples. Each sample was tested with two degranulation stimuli (fMLP and anti-FcεRI mAb) and without stimulation in duplicate (Figure S2). From this, we computed the variations in CD63+ basophil fractions resulting from decreased basophil counts. We found that these variations matched the predictions from theoretical proportion statistics, with variations declining with increasing basophil counts and absolute variations being maximal at mid-level activations (Figure 2A, Figures S3 and S4).

To further emphasize the importance of a statistical cutoff, we investigated the effect of low basophil counts on the rate of false positives, that is, the probability of recording a positive result on a negative (unstimulated) sample. In the case of fixed cutoffs, fewer than 450 basophils resulted in a heightened rate of false positives, particularly for cutoffs below 10% (Figure 2B,C). In contrast, considering cutoffs with their confidence margins imposes stricter criteria when stochastic variation is high, thereby keeping false positives to a minimum, even down to 50 basophils (Figure 2D). This analysis also applies to peanut-allergic samples to minimize false negatives (Figure S5),⁶ with further validation needed for drug allergies, where activation levels are notoriously low.⁴

In summary, we have demonstrated that a statistical approach to cutoffs helps counteract the stochastic variations in CD63+ activation caused by limited basophil counts. It is then not a matter of enough basophils, but whether their measured activation reaches statistical significance compared to the cutoff. Tests with fewer basophils are conceivable, with the risk of heightened inconclusive results; but if their activation level largely exceeds the cutoff, significant reactivity may be concluded. The statistical methodology outlined herein represents a crucial aspect for increasing clinical confidence in BAT for allergy diagnosis.

LB, TS, MS, CB and MAG were involved in conceptualisation. LB and CB were involved in formal analysis. DV, AM and ML were involved in investigation. AM, CG and CB were involved in data curation. LB was involved in writing–original draft. LB, DV, AM, CG, TS, MS, MR, CBG, MAG and CB were involved in writing–review and editing. LB was involved in visualisation. MAG and CB were involved in supervision. CBG, MAG and CB were involved in project administration.

This work was partially funded by a Swiss Accelerator innovation project supported by Innosuisse.

All authors are employees of BÜHLMANN Laboratories AG, which commercializes kits for basophil activation testing.