Predictors of severe systemic anaphylactic reactions inpatients with Hymenoptera venom allergy: Importance ofbaseline serum tryptase—a study of the European Academyof Allergology and Clinical Immunology Interest Group onInsect Venom Hypersensitivity

Franziska Ru€eff, MD,a Bernhard Przybilla, MD,a Maria Beatrice Bilo, MD,b Ulrich M€uller, MD,c Fabian Scheipl,d

Werner Aberer, MD,e Jo€elle Birnbaum, MD,f Anna Bodzenta-Lukaszyk, MD,g Floriano Bonifazi, MD,b Christoph Bucher,

MD,h Paolo Campi, MD,i Ulf Darsow, MD,j Cornelia Egger, MD,k Gabrielle Haeberli, MD,c Thomas Hawranek, MD,l

Michael K€orner, MD,m Iwona Kucharewicz, MD,g Helmut K€uchenhoff, PhD,d Roland Lang, PhD,l Oliviero Quercia, MD,n

Norbert Reider, MD,k Maurizio Severino, MD,i Michael Sticherling, MD,o Gunter Johannes Sturm, MD,e

and Brunello W€uthrich, MDh Munich, Hannover, and Erlangen, Germany, Ancona, Florence, and Faenza, Italy, Bern and Zurich,

Switzerland, Graz, Innsbruck, and Salzburg, Austria, Marseille, France, and Bialystock, Poland

Background: Severe anaphylaxis to honeybee or vespid stings isassociated with a variety of risk factors, which are poorlydefined.Objective: Our aim was to evaluate the association of baselineserum tryptase concentrations and other variables routinelyrecorded during patient evaluation with the frequency of pastsevere anaphylaxis after a field sting.Methods: In this observational multicenter study, we enrolled962 patients with established bee or vespid venom allergy whohad a systemic reaction after a field sting. Data were collectedon tryptase concentration, age, sex, culprit insect,cardiovascular medication, and the number of preceding minorsystemic reactions before the index field sting. A severe reactionwas defined as anaphylactic shock, loss of consciousness, orcardiopulmonary arrest. The index sting was defined as thehitherto first, most severe systemic field-sting reaction. Relativerates were calculated with generalized additive models.Results: Two hundred six (21.4%) patients had a severeanaphylactic reaction after a field sting. The frequency of thisevent increased significantly with higher tryptase concentrations(nonlinear association). Other factors significantly associatedwith severe reactions after a field sting were vespid venom

allergy, older age, male sex, angiotensin-converting enzymeinhibitor medication, and 1 or more preceding field stings with aless severe systemic reaction.Conclusion: In patients with honeybee or vespid venom allergy,baseline serum tryptase concentrations are associated with therisk for severe anaphylactic reactions. Preventive measuresshould include substitution of angiotensin-converting enzymeinhibitors. (J Allergy Clin Immunol 2009;124:1047-54.)

Key words: Hymenoptera venom, allergy, anaphylactic shock, tryp-tase, risk factors, angiotensin-converting enzyme inhibitor, age, sex

Systemic allergic reactions to insect stings are reported by0.3% to 7.5% of persons in the United States and Europe.1,2 Up toone fifth of these subjects will eventually experience severe life-threatening reactions requiring emergency department admis-sion.3 Estimates of the incidence of insect sting–related mortalitycaused by early anaphylaxis range between 0.03 and 0.48 fatali-ties per 1,000,000 inhabitants per year,1 but true numbers andthe incidence of near-fatal episodes might be considerablyhigher.2 Consequently, European and North American practiceguidelines recommend specific venom immunotherapy (VIT) in

From athe Department of Dermatology and Allergology, Ludwig-Maximilians-

Universitat, Munich; bthe Allergy Unit, Department of Internal Medicine, Allergy,

Immunology and Respiratory Diseases, Ospedali Riuniti di Ancona, Azienda Ospeda-

liero-Universitaria, Ancona; cAllergiestation Medizinische Klinik, Spital Ziegler,

Spitalnetz Bern; dthe Statistical Consulting Unit, Department of Statistics, Ludwig-

Maximilians-Universitat, Munich; ethe Department of Dermatology, Medical Univer-

sity of Graz; fService de Pneumo-allergologie, Hopital Sainte-Marguerite, Marseille;gthe Department of Allergology and Internal Medicine, Medical University of Bialys-

tok; hAllergiestation, Dermatologische Klinik und Poliklinik, Universitatsspital

Z€urich, Zurich; ithe Allergy Clinic, Nuovo Ospedale San Giovanni di Dio, Florence;jthe Department of Dermatology and Allergy Biederstein, Technische Universitat

M€unchen, and the Division of Environmental Dermatology and Allergy Helmholtz

Center/TUM, Munich; kthe Department of Dermatology, Medical University of Inns-

bruck; lthe Department of Dermatology, Paracelsus Private Medical University, Salz-

burg; mthe Department of Dermatology and Allergology, Hannover Medical School;nDipartimento di Medicina Interna, Ospedale per gli Infermi, Faenza; and oHautklinik,

Universitatsklinikum Erlangen.

Supported by grants for the collection of the data from Phadia (Freiburg, Germany) and

ALK-Abello (Wedel, Germany).

Disclosure of potential conflict of interest: F. Ru€eff has received lecture honoraria from

ALK-Abello, Phadia, Bencard, and HAL and has received research support from

Allergopharma and HAL. B. Przybilla has received research support from Informa-

tionsverbund Dermatologischer Kliniken and has provided legal consultation or

expert witness testimony on the topic of occupational dermatoses. M. B. Bilo has

received lecture honoraria from Stallergenes. F. Scheipl has received research support

from Deutsche Forschungsgemeinschaft. W. Aberer has received lecture honoraria

from ALK-Abello, Phadia, and Bencard. A. Bodzenta-Lukaszyk and I. Kucharewicz

have received honoraria from HAL. F. Bonifazi has received lecture honoraria from

Phadia. T. Hawranek has received lecture honoraria and research support from ALK-

Abello. N. Reider has received lecture honoraria and research support from ALK-

Abello, HAL, Bencard, and Phadia. G. J. Sturm has received lecture honoraria from

ALK-Abello. The rest of the authors have declared that they have no conflict of

interest.

Received for publication May 22, 2009; revised July 13, 2009; accepted for publication

August 11, 2009.

Reprint requests: Franziska Ru€eff, MD, Klinik und Poliklinik f€ur Dermatologie und Al-

lergologie, AllergieZentrum, Ludwig-Maximilians-Universitat M€unchen, Frauen-

lobstr. 9-11, D-80337 Munich, Germany. E-mail: Franziska.Rueff@med.

uni-muenchen.de.

0091-6749/$36.00

� 2009 American Academy of Allergy, Asthma & Immunology

doi:10.1016/j.jaci.2009.08.027

1047

J ALLERGY CLIN IMMUNOL

NOVEMBER 2009

1048 RUEFF ET AL

Abbreviations used

CE: Angiotensin-converting enzyme

BTC: Baseline tryptase concentration

EAACI: European Academy of Allergology and Clinical Immunology

GAM: Generalized additive model

ROC: Receiver operating characteristic

VIT: Venom immunotherapy

subjects with venom allergy of any age who experience respira-tory or cardiovascular symptoms to insect stings.4,5 For subjectswith systemic reactions limited to the skin, European guidelines4

recommend VIT only if additional risk factors or impairment ofquality of life are present, whereas North American guidelines5

recommend VIT for all adult patients with systemic reactions ofany kind.

In the past, numerous authors have tried to define specific riskfactors for severe systemic sting reactions, thereby improving theselection of patients for immunotherapy. One of those risk factorsmight be the individual mast cell burden. Mast cells are the keycells involved in immediate-type allergies, and it has beenproposed that the activity of mast cells, number of mast cells,or both correlate with the severity of allergic reactions.6,7 Becausemast cell burden of the body can be reflected by the serum base-line tryptase concentration (BTC), several authors have used thatvariable to characterize patients with Hymenoptera venom al-lergy.8-12 It has been repeatedly observed that patients with Hy-menoptera venom allergy and increased BTCs experience moresevere allergic reactions to field stings than patients with normallevels.10,11 Conversely, patients who present with a history of a se-vere anaphylactic reaction or who have demonstrated such majorreactions during a sting challenge are more likely to have an in-creased BTC.8,9,12 However, significant uncertainties limit theclinical relevance of the latter studies. The low frequency of se-vere systemic reactions did not allow a comprehensive and statis-tically sound analysis of other predictive variables interferingwith the tryptase effect. Furthermore, each of these studies wasperformed in a single center, and they only included a compara-tively small number of patients.

These shortcomings stimulated the European Academy ofAllergology and Clinical Immunology (EAACI) Interest Groupon Insect Venom Hypersensitivity to perform a large-scalemulticenter study, which focused on BTCs and other suspectedrisk factors for potentially life-threatening reactions after a fieldsting in unselected patients with Hymenoptera venom allergy.

METHODS

Study designThe Tryptase in Hymenoptera Venom Allergy study of the EAACI Interest

Group on Insect Venom Hypersensitivity is a partly retrospective and partly

prospective observational cohort study performed in 14 European clinics

specializing in the diagnosis and treatment of allergic diseases. The study

consists of 3 parts and evaluates patients with Hymenoptera venom allergy

who were enrolled consecutively and prospectively. Part I of the study is a

retrospective analysis addressing the importance of specific risk factors for

severe sting reactions after a field sting according to the patient’s history. Parts II

and III are prospective and observational studies evaluating enrolled patients

with respect to treatment efficacy and risk factors for systemic reactions during

immunotherapy. Results of parts II and III will be the subject of future

publications.

For part I, the primary dependent study variable was a severe systemic

reaction (anaphylactic shock) when stung by a honeybee or a vespid species in

the field. The study was designed to permit the detection of a 1.33-fold

difference in the incidence of a severe systemic reaction between 2 equal

groups according to a serum BTC of greater or less than 11.4 mg/L. At least

200 new cases of severe systemic reactions were required to provide the study

with sufficient power to detect such an increase in risk (2-sided type I error,

5%; power, 90%). Based on the data from the Hymenoptera Venom Study,13

the frequency of severe systemic reactions among patients with systemic aller-

gic reactions at a field sting was assumed to be 250 per 1,000 allergic patients,

and hence the study had to collect at least 800 patients with Hymenoptera

venom allergy before the primary objective could be examined. The prospec-

tive patient enrollment and the analysis of retrospectively collected patient

data were approved by the institutional review board of each participating cen-

ter, and each patient or patient’s parent consented to an anonymous data

analysis.

Collection of dataEnrollment. Eligible patients were consecutively enrolled in the study

from clinics in each of the participating centers beginning in May 2001;

enrollment was complete by August 2003. The whole study was terminated in

2009 after completion of the follow-up examinations necessary to evaluate

treatment efficacy. Eligible patients were those who had experienced a

systemic reaction after a preceding field sting who had a hitherto untreated

Hymenoptera venom allergy. Patients who were allergic to both bee and

vespid venom were not included in the analysis nor were patients in whom it

was not possible to unequivocally attribute clinical symptoms to a specific

insect or in whom we could not precisely determine the exact date of the last

allergic reaction.

After enrollment, 2 standardized data collection forms were completed at

the sites to provide information about the patients’ histories and the results of

testing. The Medical Advisory Committee of the study (Beatrice Bilo, Jo€elle

Birnbaum, Floriano Bonifazi, Pamela Ewan, Marek Jutel, Hannecke Oude

Elberink, Holger Mosbech, Ulrich M€uller, Bernhard Przybilla, and Franziska

Ru€eff) approved the final version of the forms (developed by F. Ru€eff and B.

Przybilla).

Patient history. The patient history form included information about

insect sting history, medication, and demographic data. The index sting, to

which all other information was referred, was defined as that field sting that

had resulted in the hitherto most severe systemic reaction according to the

individual patient’s history. If a patient had several sting reactions with a

comparable degree of severity and that were the most severe reactions overall,

the first, most severe sting was taken as the index sting.

For all stings, we requested in detail which type of insect could have

stung the subject and which kind of the following systemic symptoms had

occurred: flush; pruritus; urticaria; angioedema (not at sting site); upper or

lower airway obstruction; decrease of blood pressure; gastrointestinal

symptoms (vomiting, diarrhea, and abdominal colic); unconsciousness,

shock, or both; and requirement of antiallergic emergency therapy,

including resuscitation. We also recorded the specific environmental

conditions at the time of the sting. The judgment of the attending physician

was relied on for determining whether such symptoms as a decrease in

blood pressure, unconsciousness, or shock were believably reported by the

patient.

We furthermore recorded sex, age, and the type of medication the

patient was taking at the time of the index sting (angiotensin-converting

enzyme [ACE] inhibitors, selective or nonselective b-blockers, or any kind

of antihypertensive medication). Finally, a blood sample was taken at the

time of the first patient visit to determine BTCs and concentrations of

venom-specific IgE antibodies. At least 14 days had to have passed

between blood sampling and the time of the last allergic reaction to

exclude potential interferences between preceding allergic events and

BTCs.

Diagnostic procedures. The diagnostic procedures form de-

scribed details of diagnostic procedures, including venom-specific IgE

antibody concentrations and skin testing with venoms, and their

J ALLERGY CLIN IMMUNOL

VOLUME 124, NUMBER 5

RUEFF ET AL 1049

corresponding results. For data analysis, patients allergic to vespid genera

(Vespula, Vespa, or Polistes species) were combined into one group.

Data accuracy. The accuracy of data in the forms was ensured by the

specific qualification of research staff. Institutional supervisors monitored

local data entries by looking over 5% to 10% of the forms. All patients were

assigned a unique consecutive code number during a single clinic visit, and

coded forms were transmitted to a central depository without identification of

the patient. Each patient form was checked for completeness and plausibility

by members of the steering committee. Oversight of data collection and

analysis, integrity of the data, and research is provided by the EAACI Interest

Group.

Diagnostic principles. Diagnosis of Hymenoptera venom allergy

followed specific guidelines published by the EAACI and the American

Academy of Allergy, Asthma & Immunology.4,5,14 A skin prick test response

was read as positive at a mean wheal diameter of 3 mm at a venom concen-

tration of 100 mg/mL or less, and an intradermal test result was read as pos-

itive at a mean wheal diameter of 5 mm at a concentration of 1 mg/mL or

less. Allergy was diagnosed if a patient presented with a conclusive history,

including entomologic identification, and a corresponding venom sensitiza-

tion (according to an identification of venom-specific serum IgE, a positive

skin test reaction to venom, or both). Patients with inconsistent results or

in whom no unequivocal diagnosis could be obtained were excluded from

the analysis.

Primary dependent variable. The primary dependent variable

of the study was a severe, potentially life-threatening systemic reaction after a

field sting. Classification of a systemic reaction corresponded to that proposed

by Ring and Meßmer,15 who described 4 different grades (Table I). A severe

systemic reaction was defined as a grade III or IV reaction.

Measurement of tryptase concentration. The BTC was

determined in blood samples obtained at least 14 days after the last systemic

allergic reaction. Processing was done precisely according to the instructions

of the manufacturer of the tryptase assay. Because tryptase is quite stable and

is not released from cells usually contained in blood, we did not specify the

time allowed to pass until centrifugation. It was left to the discretion of the

individual institution when to centrifuge blood samples taken for tryptase

measurement. However, immediately after centrifugation, serum was frozen

at 2208C in each case. Each sample was marked with a code number, which

was identical with that used for the corresponding patient form. Coded

samples were shipped on dry ice to Phadia (Freiburg, Germany) for collective

measurements. Serum tryptase levels were measured by using ImmunoCAP

Tryptase. According to the manufacturer, the interassay variability for tryptase

levels between less than 1 and 100 mg/L is less than 5%. The upper 95th

percentile for healthy nonallergic subjects is 11.4 mg/L.

StatisticsCategorical variables were expressed as percentages, and continuous

variables were expressed as means 6 SDs. Comparisons between patient

groups were made by using x2 statistics or the Fisher exact test for binary var-

iables and by using t tests for continuous variables. Generalized additive

models (GAMs) were estimated by using an R package.16

TABLE I. Frequency and classification of systemic reactions

(n 5 962) at the index field sting modified according to Ring

and Meßmer15

Classification Symptoms Frequency (%)

Grade I Generalized skin symptoms (eg, flush,

generalized urticaria, angioedema)

15.2

Grade II Mild-to-moderate pulmonary,

cardiovascular, and/or gastrointestinal

symptoms

63.4

Grade III Anaphylactic shock, loss of consciousness 21.0

Grade IV Cardiac arrest, apnea 0.4

Covariate-adjusted effects of the BTC on the frequency of severe

systemic reactions at the time of the field sting were evaluated by using

multiple logistic regression models, which combined separate effects of all

individual confounding variables. For the dependent variable, we first

defined a separate starting model, which contained only the independent

variables age and sex. To construct the preliminary confounder model for

field-sting reactions, we then analyzed the following variables: type of

insect, number of preceding minor systemic reactions to the field sting, and

medication with b-blockers, ACE inhibitors, or any kind of antihypertensive

drug. Model selection for a multivariate logistic additive model was

performed by using stepwise selection based on the Akaike information

criterion. To test the tryptase effect, we then added the variable of BTC to

this preliminary GAM, thereby creating the final GAM with a nonparametric

effect for tryptase. If possible, the nonparametric tryptase term was

simplified to a log-linear term. The final GAM was then adjusted for a

potential study center effect.

The final model’s performance was evaluated by using receiver operating

characteristic (ROC) analysis. Based on the final model, an algorithm was

derived allowing the prediction of a single patient’s risk to sustain a severe

systemic reaction after a field sting. To examine whether the effect of

confounder variables, including BTC, also depends on the type of insect, we

tested interactions between insect type and the other confounder variables. For

the dependent variable, we constructed an extended final GAM, which also

included the individual interaction terms. This extended model allowed an

estimation of baseline effects in patients who had bee venom allergy on the one

hand and an estimation of the modifying effects of vespid venom allergy on the

other hand. A maximum likelihood ratio test was used to analyze the combined

effect of all interaction terms and to compare final GAMs, which did or did not

include the interactions.

RESULTS

Clinical characteristics of the cohortOne thousand forty-three subjects were originally enrolled in

the study. In compliance with the exclusion criteria, 962 hithertountreated patients were available for the evaluation of reactionsassociated with the index field sting and of their predictivevariables. Mean patient age in that cohort was 38.2 6 16.9 years;the majority of patients were male (54.4%) and had vespid venomallergy (70.0%). Some (9.9%) of the patients had already had 1 ormore preceding, less severe systemic sting reactions before theindex sting. Some (10.1%) were taking antihypertensive medica-tion at the time of the index field sting; ACE inhibitors were takenby 4.4%, and b-blockers were taken by 5.4%. Among the latter,more than 70% of the patients were receiving cardioselectiveb-blocker therapy. The mean BTC was 5.84 6 8.36 mg/L, and8.4% of the patients presented with an abnormal tryptase concen-tration (�11.4 mg/L) at the first office visit. The median time thatpassed between the index sting and measurement of the tryptaseconcentration was 3 months (range, 0.5-415 months). The mediantime that passed between the last sting reaction (which did notnecessarily have to be the reaction of the index sting, see theMethods section) and measurement of tryptase concentration was3 months (range, 0.5-364 months). The frequency at whichdifferent degrees of systemic reactions occurred at the field stingis presented in Table I. In 21.4% of the patients, the reaction to theindex sting was severe (grade III or IV).

Risk factors for severe anaphylactic reactions after a

field stingSevere systemic reactions were significantly more common in

older and male patients and also in those patients who were taking

J ALLERGY CLIN IMMUNOL

NOVEMBER 2009

1050 RUEFF ET AL

antihypertensive medication, including b-blockers and ACEinhibitors; who had vespid venom allergy; or who had 1 ormore preceding, less severe systemic sting reactions before theindex sting (Table II). After adjusting for confounder variables,we found a significant, nonlinear, continuous association betweenthe BTC and the frequency of a severe systemic sting reaction(grade III or IV, Fig 1). Other independent predictors for a severeallergic reaction were vespid venom allergy; 1 or more preceding,less severe systemic sting reactions before the index sting; malesex; and therapy with ACE inhibitors. In the multivariate analysisb-blockers were not identified as independent risk factors. Therewas also a linear association between age and the frequency of se-vere systemic symptoms (Table III).

The analysis of potential interactions between the type of insectand various confounder variables did not suggest that the mag-nitude of the tryptase effect was different in different insectspecies (P 5 .970 for the tryptase/species interaction in the ex-tended model, which included all interactions) or that the com-bined effect of all independent covariables was different inpatients with honeybee or vespid venom allergy (P 5 .680 forthe comparison between the extended model and the model with-out interactions).

Risk prediction for severe anaphylactic reactions

after a field stingAnalysis of our data revealed that tryptase cutoff concentra-

tions with respect to predicting a grade III/IV sting reaction weredependant on confounders (eg, sex and type of insect). Thereforewe used the final multivariate model, which included all con-founders, for risk prediction. This model was able to distinguishgrade I and II sting reactions from grade III and IV reactions witha classification performance of an area under the curve of 0.731.The corresponding global ROC graph is presented in Fig 2. On anindividual basis, the following procedure, based on the finalmodel, can be used to calculate a patient’s risk score and to predicthis or her risk for a grade III or IV reaction.

1. Calculate the raw risk score before taking into account theeffect of tryptase:

score 5 2 2:8 1 0:43 � xInsect11:61 � xSting2 0:63 � xGender

1 0:68 � xMedication1 0:028 � xAge;

with xInsect defined as honeybee venom allergy 5 0 and vespidvenom allergy 5 1; xSting defined as no less severe systemicsting reactions before index sting 5 0 and 1 or more less severesystemic sting reactions before index sting 5 1; xGender definedas male sex 5 0 and female sex 5 1; xMedication defined as noACE inhibitor medication at the time of index sting 5 0and ACE inhibitor medication at the time of index sting 5 1;and xage defined as age (in years).

2. Look up the multiplicative effect CTrypof the patient’s BTC inFig 3 and multiply escore with it to get the estimated odds for agrade III or IV reaction as follows: odds 5 escore � CTryp.

TABLE II. Distribution of the severity grade of systemic anaphylactic reactions (grade I/II or III/IV) after the index sting with respect to

baseline parameters

Parameter

Grade I or II

reaction (n 5 756)

Grade III or IV

reaction (n 5 206) P value

b-Blocker medication at the time of the index sting Yes 34 (65.4%) 18 (34.6%) .024

No 722 (79.3%) 188 (20.7%)

ACE inhibitor medication at the time of the index sting Yes 24 (57.1%) 18 (42.9%) .002

No 732 (79.6%) 188 (20.4%)

Any antihypertensive medication at the time of the index sting Yes 61 (63.5%) 36 (36.5%) <.001

No 695 (80.4%) 170 (19.6%)

Sex Male 385 (73.6%) 138 (26.4%) <.001

Female 371 (84.5%) 68 (15.5%)

One or more preceding, less severe systemic sting reactions before index sting Yes 46 (48.4%) 49 (51.6%) <.001

No 710 (81.9%) 157 (18.1%)

Insect responsible for index sting and associated allergic reaction Bee 241 (83.4%) 48 (16.6%) .016

Vespid 515 (76.5%) 158 (23.5%)

Age (y) at index sting according to median <38 424 (86.2%) 68 (13.8%) <.001

�38 332 (70.6%) 138 (29.4%)

Associations are shown between clinical, demographic, and therapeutic parameters and the sting reaction that caused the most severe reaction before first office visit. Percentages

indicate the frequency of a certain degree of sting reaction referred to the total number of patients presenting with a specific parameter value. The P value refers to the univariate

analysis.

FIG 1. Smoothed function and 95% confidence band (dashed lines) for the

effect of BTCs on the risk of a severe allergic reaction (grade III or IV) after

the index field sting (final multivariate model). Odds ratios refer to those of

the median of tryptase concentration. The odds ratio of the latter has been

set at 1.

J ALLERGY CLIN IMMUNOL

VOLUME 124, NUMBER 5

RUEFF ET AL 1051

TABLE III. Results of the final GAM for the risk of a severe anaphylactic reaction (grade III/IV) to the index sting

Variable P value Odds ratio 95% CI

Index sting by vespid species .008 1.730 1.147 2.607

One or more preceding, less severe systemic sting reactions before the index sting <.001 4.687 2.913 7.542

Female sex <.001 0.553 0.387 0.791

ACE inhibitor medication at the time of the index sting .019 2.269 1.129 4.558

Age at the time of the index sting (per year) <.001 1.029 1.018 1.041

BTC at first office visit NA (smoothed function, see Fig 1)

Variables shown were selected according to the modeling procedure. The index sting is defined as that field sting that caused the most severe reaction before the first office visit.

NA, Not applicable.

3. Calculate the predicted risk as follows:

p 5odds

1 1 odds:

According to p, a specific predicted risk could be attributed toeach patient. With regard to sensitivity, the following cutoffvalues for the predicted risk p could be derived from the ROCof the final multivariate model: sensitivity of approximately80% 5 0.148, sensitivity of approximately 90% 5 0.0971, andsensitivity of approximately 95% 5 0.0728. However, corre-sponding specificity was consistently low (53.2%, 27.0%, and13.8%, respectively).

DISCUSSIONOur study is the first to evaluate the importance of BTCs in

the serum and of a variety of other suspected risk factors for past

FIG 2. ROC curve for the final multivariate GAM predicting the risk of a

severe allergic reaction (grade III or IV) after the index field sting.

severe systemic reactions after a field sting in untreated subjects.The only other epidemiologic study to which our study can becompared is the North American Hymenoptera Venom Study,which was purely descriptive, could not consider tryptaseconcentrations, and might not have been able to exactly diag-nose the type of insect allergy because of diagnostic shortcom-ings almost 30 years ago.13 Nevertheless, several crudeepidemiologic findings (eg, the frequency of severe systemic re-actions to the index sting [21.4% vs 25.3%]; the frequency ofpreceding, less severe systemic sting reactions before the indexsting [9.9% vs 8.0%]; or male subject predominance) are re-markably similar.

The key finding of our study is that the BTC correlatessignificantly with the risk for past, potentially life-threateningallergic sting reactions. Besides being an independent prognosticvariable, even a small increase of the BTC might be relevant forthe severity of a systemic sting reaction. It can be estimated fromthe smoothed function (Fig 1) that in untreated subjects an in-crease in tryptase concentration from a value of 4.25 mg/L (me-dian of our cohort) to 20 mg/L will simultaneously increase therisk of a severe systemic reaction after a field sting approxi-mately by a factor of 3.8. Furthermore, the association betweentryptase concentration and the extent of secondary systemic re-actions did not depend on the type of insect allergy (vespid orhoneybee).

Tryptase concentrations detectable in the circulation aremainly derived from the genes that encode a- and b-tryptases.These types of tryptases are initially produced as proenzymes(a-protryptase and b-protryptase). b-Protryptase but not a-pro-tryptase gets processed to mature b-tryptase and resides in thegranules. On appropriate stimulation (eg, during an anaphylacticreaction), active b-tryptase is released from the granules of mastcells to produce a variety of biochemical reactions, of which thebiologic significance is still uncertain.17 Without stimulation,mast cells secrete enzymatically inactive a-protryptase andb-protryptase spontaneously. It is now thought that baseline im-munoreactive tryptase in the serum (as measured by the assayused) mainly consists of those protryptases and only of smallamounts of b-tryptase.7

Originally, the BTC was used to demonstrate mast cell burdenin mastocytosis.18 Increased BTCs on the basis of an increasedmast cell number or activity can be also found in patients withother diseases or pathologies involving mast cells, such as chronicurticaria,19 uremic pruritus,20 myeloid malignancies,21 hypereo-sinophilic syndrome associated with the FIP1L1-PDGFRA fusionmutation,22 myelodysplastic syndromes,23 or mast cell activationsyndrome.24 Because mastocytosis is an established risk factor forsevere allergic reactions, increased BTCs have also been thoughtto indicate this particularly high risk.7,25,26

J ALLERGY CLIN IMMUNOL

NOVEMBER 2009

1052 RUEFF ET AL

Today there is also evidence that, apart from mastocytosis andother specific diseases, mast cell numbers and lifespan might bechronically increased in unselected allergic patients, even withoutrepeated antigen exposure.27 This observation has stimulated theconcept of screening allergic patients for an increased BTC,thereby improving the identification of high-risk patients. How-ever, it has been controversial whether minor increases in BTCare of clinical relevance.

With respect to reactions after a field sting, the currentlyaccepted upper limit for an apparently normal tryptase concen-tration (11.4 mg/L) might be inadequate. Fig 1 reveals that theslope of the odds ratio graph increases markedly above a concen-tration of approximately 5 mg/L. Consequently, it might be usefulto include the whole range of tryptase concentrations in the deci-sion of which patient should be offered immunotherapy. Measure-ment of the BTC might be especially helpful in those patients inwhom diagnosis or indication for therapy is controversial (eg, pa-tients who are fairly young, who have only mild systemic reac-tions, or in whom venom-specific IgE cannot be detected inserum by means of routine methods and in whom skin test resultsare simultaneously negative). Furthermore, measurement ofBTCs in patients with Hymenoptera venom allergy might helpidentify those who simultaneously have mastocytosis or a mono-clonal mast cell activation syndrome.26

The accuracy of the BTC concentration for precisely predictinggrade III or IV sting reactions depends on confounders, such as

FIG 3. Correction factor for BTCs. A correction factor is required to account

for the patient’s tryptase concentration to calculate the predicted risk, p, for

a severe anaphylactic reaction in an individual patient (see the Methods

section). The correction factor can be derived manually by determining

the intersection of the tryptase concentration with the graph.

sex or type of insect. Therefore we used the final multivariatemodel, which also included tryptase concentration, to derive analgorithm allowing the calculation of an individual patient’s riskof a severe systemic reaction after a field sting. Thereby cutoffscould be derived and a specific predicted risk could be attributedto each patient according to her or his characteristics. However, itshould be noted that the predictive performance of the final modelwas comparatively poor (area under the curve <0.8), most likelyresulting from other, hitherto unrecognized predictors of severeallergic reactions impossible to incorporate into our study. Onepotential candidate is general immunologic reactivity, whichmight have been variable at the time of the index sting because ofphysical activity, concurrent infections, additional temporarymedication, or alcohol consumption.

Several other important conclusions can be derived from ourresults. A particularly remarkable finding was the observation thatvespid venom allergy was an independent predictor of a higherrisk at the index field sting. Based on observations made at in-hospital insect sting challenges before treatment28 or during im-munotherapy,28,29 it has been thought that bee venom allergy isgenerally more dangerous than vespid venom allergy. However,our results suggest this association to be untrue for field stings.The only 2 other large, albeit only descriptive, studies to whichour results can be compared also found that severe, potentiallylife-threatening field-sting reactions were significantly morecommon in patients with vespid than with honeybee venomallergy.13,30

Furthermore, we found evidence for a significant associationbetween the degree of an earlier systemic sting reaction and theseverity of a subsequent sting reaction. One or more precedingstings with a less severe systemic reaction were an independentdeterminant for a severe anaphylactic reaction at the index sting.A crude association between moderate prior sting reactions and anear-fatal subsequent reaction had already been described in theHymenoptera venom study13 and in a large Australian single-cen-ter study.30

To explain this association, we have to consider the possibilityof recall bias, meaning that certain conditions might influencepatients to inflate the importance of a previous event related to thelater diagnosed disease.31 We cannot definitely exclude such a re-call bias in the context of a severe anaphylactic reaction at thetime of the index sting. However, patients with a more severe stingreaction do not seem to have a different subjective perception oftheir disease than patients with milder reactions,32 and there is noevidence that prior, less severe reactions are preconceived by thepatient as a risk factor of subsequent, more severe sting reactions.

A more likely explanation for the observed association mightbe a booster effect. Numerous experimental and clinical studieshave demonstrated corresponding booster effects in other allergicdiseases, such as asthma.5,33,34 It could also be shown that re-peated exposure to bee venom allergen accelerates systemic IgEproduction in sensitized mice.35 Therefore it is highly likelythat such boosting effects also occur in Hymenoptera venom al-lergy. Consequently, patients who have already sustained a minorsystemic reaction at a field sting might be preferred candidates fortherapy.

Another important finding was the strictly linear associationbetween age and the frequency of severe systemic reactions to theindex sting. Furthermore, medication with ACE inhibitors was theonly independent pharmacologic predictor for an increased risk.The latter finding does not support the hypothesis that general

J ALLERGY CLIN IMMUNOL

VOLUME 124, NUMBER 5

RUEFF ET AL 1053

vascular comorbidities, such as hypertension, are particularlyimportant for severe systemic reactions. According to such ahypothesis, one would expect an association of not only one but ofall types of antihypertensive medication with severe reactions.Specific cardiac comorbidities might be more important, alsoexplaining the linear effect of age. Age is a powerful predictor ofsevere adverse events after acute coronary syndromes, and afteraccounting for other factors, the odds for in-hospital deathincrease linearly by 70% for each 10-year increase in age.36

The effect of ACE inhibitors, which was independent of age,suggests a specific role for the kinin-angiotensin system withinthe mechanisms causing allergic shock. Our results provide anargument for the substitution of ACE inhibitor medication inuntreated patients with Hymenoptera venom allergy. In contrastto ACE inhibitors, we were unable to identify b-blocker therapyas an independent predictor of a severe sting reaction. Results ofprevious studies addressing the effect of concomitant b-blockermedication are conflicting37,38 and difficult to interpret because ofvarious methodological weaknesses inherent in those studies (in-sufficient number of severe anaphylactic reactions; different typesof elicitors, including b-blockers; and unequal distribution of ad-ditional risk factors). However, it should be noted that our studywas not specifically designed to evaluate the importance ofb-blocker therapy for severe sting reactions. Only 5.4% of the pa-tients were on such a therapy at the time of a field sting, and pa-tients were treated with different types of b-blockers. Thereforeour negative results do not entirely exclude the possibility thatb-blockers might have been important, although the effect is pre-sumably of much smaller magnitude than that of other predictors,such as ACE inhibitor medication or tryptase concentration.

Also, male sex was an independent risk factor for severesystemic reactions to the index field sting. This association is incontrast to other immediate-type allergies, in which sex hormonesappear to increase the risk for anaphylactic reactions in femalesubjects.39 The effect of male sex in insect sting–induced anaphy-laxis presumably results from a selection effect. Because of a dif-ferent degree of exposure, adult men are stung more frequentlythan women and might therefore be at higher risk for sensitizationor severe allergic reactions.13

We conclude that in patients with Hymenoptera venom allergy,even minor increases in BTCs are associated with more frequentsevere systemic reactions after a field sting, independently ofother prognostic variables. Older patients and those who havealready sustained a minor allergic reaction after a field sting mightbe preferred candidates for immunotherapy. For prevention, asubstitution of ACE inhibitors should be considered.

We thank W. Hartl, MD, for the critical revision of the manuscript.

Clinical implications: In patients with Hymenoptera venom al-lergy, the severity of anaphylactic sting reactions varies withBTCs. The latter might be included in the decision of which pa-tients should be offered immunotherapy.

REFERENCES

1. Bilo BM, Bonifazi F. Epidemiology of insect-venom anaphylaxis. Curr Opin Al-

lergy Clin Immunol 2008;8:330-7.

2. Graft DF. Insect sting allergy. Med Clin North Am 2006;90:211-32.

3. Clark S, Long AA, Gaeta TJ, Camargo CA Jr. Multicenter study of emergency de-

partment visits for insect sting allergies. J Allergy Clin Immunol 2005;116:643-9.

4. Bonifazi F, Jutel M, Bilo BM, Birnbaum J, Muller U. EAACI Interest Group on

Insect Venom Hypersensitivity. Prevention and treatment of hymenoptera venom

allergy: guidelines for clinical practice. Allergy 2005;60:1459-70.

5. Joint Task Force on Practice Parameters; American Academy of Allergy. Asthma

and Immunology; American College of Allergy, Asthma and Immunology; Joint

Council of Allergy, Asthma and Immunology. Allergen immunotherapy: a practice

parameter second update. J Allergy Clin Immunol 2007;120(suppl):S25-85.

6. Ru€eff F, Placzek M, Przybilla B. Mastocytosis and Hymenoptera venom allergy.

Curr Opin Allergy Clin Immunol 2006;6:284-8.

7. Schwartz LB. Diagnostic value of tryptase in anaphylaxis and mastocytosis. Immu-

nol Allergy Clin North Am 2006;26:451-63.

8. Haeberli G, Br€onnimann M, Hunziker T, M€uller U. Elevated basal serum tryptase

and hymenoptera venom allergy: relation to severity of sting reactions and to safety

and efficacy of venom immunotherapy. Clin Exp Allergy 2003;33:1216-20.

9. Schwartz LB, Bradford TR, Rouse C, Irani AM, Rasp G, Van der Zwan JK, et al.

Development of a new, more sensitive immunoassay for human tryptase: use in

systemic anaphylaxis. J Clin Immunol 1994;14:190-204.

10. Ludolph-Hauser D, Ru€eff F, Fries C, Sch€opf P, Przybilla B. Constitutively raised

serum concentrations of mast-cell tryptase and severe anaphylactic reactions to Hy-

menoptera stings. Lancet 2001;357:361-2.

11. Sturm GJ, B€ohm E, Trummer M, Weiglhofer I, Heinemann A, Aberer W. The

CD63 basophil activation test in Hymenoptera venom allergy: a prospective study.

Allergy 2004;59:1110-7.

12. Kucharewicz I, Bodzenta-Lukaszyk A, Szymanski W, Mroczko B, Szmitkowski M.

Basal serum tryptase level correlates with severity of hymenoptera sting and age.

J Investig Allergol Clin Immunol 2007;17:65-9.

13. Lockey RF, Turkeltaub PC, Baird-Warren IA, Olive CA, Olive ES, Peppe BC, et al.

The Hymenoptera venom study I, 1979-1982: demographics and history—sting

data. J Allergy Clin Immunol 1988;82:370-81.

14. Bilo BM, Ru€eff F, Mosbech H, Bonifazi F, Oude-Elberink JN. the EAACI Interest

Group on Insect Venom Hypersensitivity. Diagnosis of Hymenoptera venom al-

lergy. Allergy 2005;60:1339-49.

15. Ring J, Meßmer K. Incidence and severity of anaphylactoid reactions to colloid

volume substitutes. Lancet 1977;1:466-9.

16. R Development Core Team. R: a language and environment for statistical comput-

ing. R Foundation for Statistical Computing 2009. Available athttp://www.R-

project.org Accessed April 9, 2009.

17. Hallgren J, Pejler G. Biology of mast cell tryptase. An inflammatory mediator.

FEBS J 2006;273:1871-95.

18. Schwartz LB, Sakai K, Bradford TR, Ren S, Zweiman B, Worobec AS, et al. The alpha

form of human tryptase is the predominant type present in blood at baseline in normal

subjects and is elevated in those with systemic mastocytosis. J Clin Invest 1995;96:2702-10.

19. Hidvegi B, Nagy E, Szabo T, Temesvari E, Marschalko M, Karpati S, et al. Corre-

lation between T-cell and mast cell activity in patients with chronic urticaria. Int

Arch Allergy Immunol 2003;132:177-82.

20. Dugas-Breit S, Sch€opf P, Dugas M, Schiffl H, Ru€eff F, Przybilla B. Baseline serum

levels of mast cell tryptase are raised in hemodialysis patients and associated with

severity of pruritus. J Dtsch Dermatol Ges 2005;3:343-7.

21. Sperr WR, Hauswirth AW, Valent P. Tryptase a novel biochemical marker of acute

myeloid leukemia. Leuk Lymphoma 2002;43:2257-61.

22. Klion AD, Noel P, Akin C, Law MA, Gilliland DG, Cools J, et al. Elevated serum

tryptase levels identify a subset of patients with a myeloproliferative variant of id-

iopathic hypereosinophilic syndrome associated with tissue fibrosis, poor progno-

sis, and imatinib responsiveness. Blood 2003;101:4660-6.

23. Sperr WR, Stehberger B, Wimazal F, Baghestanian M, Schwartz LB, Kundi M,

et al. Serum tryptase measurements in patients with myelodysplastic syndromes.

Leuk Lymphoma 2002;43:1097-105.

24. Akin C, Scott LM, Kocabas CN, Kushnir-Sukhov N, Brittain E, Noel P, et al. Dem-

onstration of an aberrant mast-cell population with clonal markers in a subset of

patients with ‘‘idiopathic’’ anaphylaxis. Blood 2007;110:2331-3.

25. Greenhawt M, Akin C. Mastocytosis and allergy. Curr Opin Allergy Clin Immunol

2007;7:387-92.

26. Bonadonna P, Perbellini O, Passalacqua G, Caruso B, Colarossi S, Dal Fior D, et al.

Clonal mast cell disorders in patients with systemic reactions to Hymenoptera stings

and increased serum tryptase levels. J Allergy Clin Immunol 2009;123:680-6.

27. Kawakami T, Kitaura J. Mast cell survival and activation by IgE in the absence of

antigen: a consideration of the biologic mechanisms and relevance. J Immunol

2005;175:4167-73.

28. Ru€eff F, Przybilla B, M€uller U, Mosbech H. The sting challenge test in Hymenoptera

venom allergy. Position paper of the Subcommittee on Insect Venom Allergy of the Eu-

ropean Academy of Allergology and Clinical Immunology. Allergy 1996;51:216-25.

29. M€uller U, Helbling A, Berchtold E. Immunotherapy with honeybee venom and yel-

low jacket venom is different regarding efficacy and safety. J Allergy Clin Immunol

1992;89:529-35.

J ALLERGY CLIN IMMUNOL

NOVEMBER 2009

1054 RUEFF ET AL

30. Solley GO. Stinging and biting insect allergy: an Australian experience. Ann Al-

lergy Asthma Immunol 2004;93:532-7.

31. Hassan E. Recall Bias can be a threat to retrospective and prospective research

designs. Internet J Epidemiol. Available at: http://www.ispub.com/journal/the_

internet_journal_of_epidemiology/archive/volume_3_number_2_12.html.

32. Oude Elberink JN, Dubois AE. Quality of life in insect venom allergic patients.

Curr Opin Allergy Clin Immunol 2003;3:287-93.

33. Smith N, Broadley KJ. Optimisation of the sensitisation conditions for an ovalbu-

min challenge model of asthma. Int Immunopharmacol 2007;7:183-90.

34. Takeda H, Kogame A, Tanaka H, Nagai H. Time course study for airway inflam-

mation and responsiveness by repeated provocation of aeroantigen in guinea pigs.

Prostaglandins 1997;54:805-20.

35. Okano M, Nishizaki K, Satoskar AR, Yoshino T, Masuda Y, Harn DA Jr. Involve-

ment of carbohydrate on phospholipase A2, a bee-venom allergen, in in vivo anti-

gen-specific IgE synthesis in mice. Allergy 1999;54:811-8.

36. Granger CB, Goldberg RJ, Dabbous O, Pieper KS, Eagle KA, Cannon CP, et al. Global

Registry of Acute Coronary Events Investigators. Predictors of hospital mortality in the

Global Registry of Acute Coronary Events. Arch Intern Med 2003;163:2345-53.

37. Brown SG. Clinical features and severity grading of anaphylaxis. J Allergy Clin

Immunol 2004;114:371-6.

38. Lang DM. Do beta-blockers really enhance the risk of anaphylaxis during immu-

notherapy? Curr Allergy Asthma Rep 2008;8:37-44.

39. Chen W, Mempel M, Schober W, Behrendt H, Ring J. Gender difference, sex

hormones, and immediate type hypersensitivity reactions. Allergy 2008;63:1418-27.