Systematic review of prevalence of aspirin induced asthma and its implications for clinical practice

BMJ 2004; 328 doi: (Published 19 February 2004) Cite this as: BMJ 2004;328:434
  1. Christine Jenkins, head, asthma group (crj{at},
  2. John Costello, clinical director of medicine2,
  3. Linda Hodge, dietitian3
  1. 1Woolcock Institute of Medical Research, Royal Prince Alfred Hospital, Camperdown, NSW 2050, Australia
  2. 2King's College Hospital, London SE5 9RS
  3. 3Asthma Centre, Royal Prince Alfred Hospital
  1. Correspondence to: C Jenkins, Suite 702, 26 Ridge Street, North Sydney, NSW
  • Accepted 14 November 2003


Objective To reassess the prevalence of aspirin induced asthma and other issues related to the syndrome.

Data sources Biosis, SciSearch (1990 to March 2002), Embase (1974 to March 2002), Medline (1966 to March 2002), Toxline, Derwent Drug File (1964 to March 2002), Conference Papers Index and Inside Conferences, Int'l Pharmaceutical Abstracts, Pharma-Online (1978 to March 2002).

Selection criteria Study type, patient population, and outcome measures. Review was restricted to respiratory responses to analgesics available without prescription.

Results The prevalence of aspirin induced asthma was highest when determined by oral provocation testing (adults 21%, 95% confidence interval 14% to 29%; children 5%, 0% to 14%) than by verbal history (adults 3%, 2% to 4%; children 2%, 1% to 3%). Cross sensitivity to doses of over the counter non-steroidal anti-inflammatory drugs was present in most patients with aspirin induced asthma: ibuprofen, 98%; naproxen, 100%; and diclofenac, 93%. The incidence of cross sensitivity to paracetamol among such patients was only 7%.

Conclusions Aspirin induced asthma in adults is more prevalent than previously suggested. When there is a clinical necessity to use aspirin or a non-steroidal anti-inflammatory drug and there is uncertainty about safety, oral provocation testing should be performed.


Aspirin induced asthma is a distinct clinical syndrome affecting some asthmatic patients. It is characterised by the onset of asthma 30 minutes to three hours after the ingestion of aspirin. Although the name of the condition relates to aspirin, it is well established that affected patients are cross sensitive to all non-steroidal anti-inflammatory drugs (NSAIDs) that inhibit cyclo-oxygenase (COX) enzymes.1 2 Paracetamol (acetaminophen), however, is seldom associated with cross sensitivity in patients with aspirin induced asthma. Emerging evidence shows that paracetamol may exert at least part of its analgesic effect through a newly identified COX-3 isoenzyme, whereas aspirin induced asthma is believed to involve inhibition of COX-1.35

Despite a wealth of literature on aspirin induced asthma, controversy remains as to its prevalence, with published data ranging from 4% to 44%. Differences in populations studied, methods used,definitions of outcomes, and criteria for defining sensitivity reactions may all be responsible for the variations in reported rates.68 A greater understanding of aspirin induced asthma is desirable,particularly given the increasing trend for consumers to treat themselves for minor painful conditions and the lack of association by many consumers between asthma and some analgesics. We performed a systematic review to reassess the prevalence of aspirin induced asthma in the general asthma population and to understand better the cross sensitivity of these individuals to commonly used non-prescription analgesics.


On 3 March 2002 we identified articles, in any language, with data on aspirin sensitivity among asthmatic patients and the use of paracetamol or NSAIDs. Additional articles were found through archives and the reference lists of identified articles. We excluded from the main analysis studies reporting non-respiratory responses to analgesics, such as urticaria.

Analysis of prevalence

Most of the studies recruited from asthma clinics or hospitals where patients had presented with acute exacerbations. To account for preselection bias, we subdivided the participants into threegroups: group 1, all patients with asthma—with or without a history of aspirin induced asthma and with or without markers of an increased risk or likelihood of the syndrome; group 2, patients preselected on the basis that they had either a reliable history of aspirin induced asthma or markers of an increased risk or likelihood of the syndrome; and group 3, patients with no markers of an increased risk or likelihood of aspirin induced asthma.

The primary outcome was to determine whether the ingestion of aspirin triggered an asthmatic response. For this reason we included in the main analysis only studies in which patients underwent provocation challenges. Our analysis defined a positive aspirin induced asthma response as a 20% or more reduction in forced expiratory volume in one second within three or four hours of the challenge, as recently described.9

Owing to the potential for severe reactions in patients with aspirin induced asthma, provocation challenges in patients with an unequivocal history of aspirin sensitivity is deemed unethical. In view of this, many authors had combined the results from such patients with the number of patients showing a positive reaction to provocation challenge; we included such studies in a subanalysis.

The use of history alone for determining aspirin sensitivity among asthmatic patients has underestimated and overestimated prevalence. Therefore we also conducted a subanalysis of papers in which history was the only means of determining aspirin sensitivity.

Analysis of incidence of cross sensitivity to NSAIDs or paracetamol

In the analysis of cross sensitivity we included only level 1 studies representing properly controlled, randomised, and single blinded or double blinded clinical trials. The primary outcome wasto determine whether the ingestion of NSAIDs (specifically ibuprofen, naproxen, and diclofenac) orparacetamol triggered an asthmatic response in patients who had been positively identified as having aspirin induced asthma by oral provocation testing, history, or both. Studies were only included for analysis if cross sensitivity to NSAIDs or paracetamol was determined by provocation challenge.

Statistical analysis

Using a weighted average of the incidence rates from individual studies, we calculated pooled incidence rates and 95% confidence intervals. The reciprocal of the variance in each study was calculated for weighting.


Prevalence in adults

A total of 66 papers were identified that gave the prevalence for aspirin induced asthma. Only 21 (15 in adults and six in children) were eligible for inclusion in our analysis (fig 1).8 1029 Although a double blind trial would produce more robust data, this is not the usual method employed for studies of aspirin induced asthma owing to the high risk of life threatening reactions. Only four of the trials were double blind.

The pooled incidence of aspirin induced asthma was 21% (95% confidence interval 14% to 29%), regardless of whether the patients had a history of aspirin induced asthma or markers for an increased risk of the syndrome (table 1). Prevalence of aspirin induced asthma also seemed to depend on the method used to determine it, with history alone resulting in a much lower prevalence (2.7%). Four of the studies in adults gave data on the number of patients reacting to different doses of aspirin.1214 1517 Around half (57/113) of those who had positive reactions, did so at low doses of aspirin (≥80 mg), indicating that they were highly sensitive.

Table 1

Prevalence of aspirin induced asthma in adults, analysed by population and test method. Values are numbers (percentages) unless stated otherwise

View this table:

Prevalence in children

Aspirin induced asthma has been considered rare in children, yet we found that although it is less common in children than in adults, prevalence is still around 5% (0% to 14%) when children are subject to oral provocation testing (table 2).21 23 25 Although only one of the studies was a double blind, randomised controlled trial, it accounted for almost half of the patients in our analysis.21 As with adults, the use of history alone gave a lower estimate of prevalence (2%, 1% to3%) than determined by oral provocation testing.

Table 2

Prevalence of aspirin induced asthma in children, analysed by population and test method. Values are numbers (percentages) unless stated otherwise

View this table:

Incidence of cross sensitivity

Over the counter NSAIDs

Ten studies reported the incidence of cross sensitivity to three commonly used NSAIDs (ibuprofen, naproxen, diclofenac). Only three of these were level 1 studies eligible for inclusion.1 30 31 Based on these, the incidence of cross sensitivity was: ≤ 400 mg ibuprofen, 98% (95% CI 90% to 100%); ≤100 mg naproxen, 100% (83% to 100); and ≤ 40 mg diclofenac, 93% (76% to 100%; table 3).130 31

Table 3

Incidence of cross sensitivity to NSAIDs among patients with aspirin induced asthma

View this table:

Cross sensitivity to paracetamol

Each article on cross sensitivity to paracetamol was classified according to its methods. Ten of 52 identified papers were of level 1 studies.1 1214 17 24 3236 Table 4 summarises the data for these studies, except for one in which the authors report the number of oral provocation tests and reactions but not the number of patients with a positive reaction.36 Of 268 adults and children with aspirin induced asthma who underwent oral challenge, only 32 had a positive respiratory reaction to paracetamol (pooled incidence 7%, 0% to 16%).

Table 4

Incidence of cross sensitivity to paracetamol among patients with aspirin induced asthma

View this table:


The prevalence of aspirin induced asthma is 21% for adults and 5% for children according to our systematic review. Our review is, however, limited by the retrospective nature of the analysis and the heterogeneity of the patient population. We were unable to attain the clinical status of all the patients, there was a lack of uniformity with the challenge tests, and some studies included only a small number of patients. Clinical heterogeneity was overcome by analysing different patient populations separately, and the findings are strengthened by the distinction between adults and children and the types of testing involved (oral provocation testing, patient history, questionnaire). Prospective studies in the general asthma population—correlating questionnaire responses withthe results of provocation challenge—are warranted to confirm the prevalence of the data we have derived.

Our value for the prevalence of aspirin induced asthma in adult asthmatic patients is higher than the 10% reported in recent reviews.7 37 Methods and patient bias provide one explanation for this, since studies assessing patient history only resulted in lower estimates than those based on oral provocation testing. It is not surprising that asthmatic patients often do not report sensitivity to aspirin when questioned because many are unaware of this syndrome and do not associate the use of a pain reliever with an asthma attack. Indeed, 15% of asthmatic patients evaluated during a multicentre study spanning 10 European countries only became aware of their intolerance after provocation testing.38 Under-diagnosis of the syndrome may be due to the lack ofroutine testing by aspirin challenge in asthmatic patients who do not report a positive history ofaspirin sensitivity.38

Analyses based on the use of a questionnaire resulted in a higher number of positive results than did retrospective analyses of medical records. Prevalence rates of 11-24% were given in the four studies using questionnaires,8 16 18 20whereas rates of 2-3% were obtained from the three studies relying on medical records.26 28 One explanation for this finding is that questionnaires are usually deployed within a set time frame and by a limited numberof people, whereas medical records are likely to be completed by a larger number of different healthcare professionals—the absence of a recording of aspirin induced asthma does not necessarily mean that it was not present. Retrospective analysis of medical records is therefore an insensitive means of detecting aspirin induced asthma.

Cross sensitivity to NSAIDs occurred in most of the patients with aspirin induced asthma. In contrast, the pooled incidence of cross sensitivity to paracetamol among patients with the syndrome was only 7%. Based on these data, less than 2% of asthmatic patients are likely to react to both paracetamol and aspirin. The available data indicate that patients who are highly sensitive to aspirin are more likely to be sensitive to paracetamol than those requiring higher doses of aspirin toelicit a response. The authors of one study pointed out that their patients were highly sensitive to aspirin—the mean provoking dose was only 47 mg of aspirin, compared with 1227 mg of paracetamol.33 They also correlated aspirin dose and frequency of cross reactivity to paracetamol (fig 2), commenting that when the aspirin dose provoking a 20% reduction in forced expiratory volume in one second was 30 mg, 83%(5/6) of patients reacted to paracetamol, but when the dose was 150 mg, none (0/4) of the patientsreacted to paracetamol. This study was also used to compare the severity of paracetamol induced bronchoconstriction with corresponding reactions to aspirin in the same patients.33 Although there was no significant difference in the magnitude of the reduction in forced expiratory volume in one second with either product, the reaction to paracetamol was significantly shorter than that to aspirin (fig 3) and significantly milder, as shown by the mean number of nebuliser treatments required by the patients reacting to paracetamol (1.2 v 2.6, P = 0.035).

Fig 2
Fig 2

Relation between aspirin provoking dose and frequency of cross sensitivity to paracetamol (reproduced from Settipane et al 199533 with permission of Mosby)

Fig 3
Fig 3

Duration of aspirin induced and paracetamol induced bronchospastic reactions (reproduced from Settipane et al 199533 with permission of Mosby)

The continuing recommendation of paracetamol as the analgesic and antipyretic of first choice for patients with asthma seems warranted given the relatively low incidence of sensitivity. The newgeneration of COX-2 specific analgesics may also be safer than NSAIDs and aspirin in asthmatic patients, but further experience with these compounds is required. Based on this conclusion, we have simplified guidelines for the use of analgesics in asthmatic patients (table 5). Where history neither supports nor excludes aspirin induced asthma, and aspirin or NSAIDs are clinically indicated, formal provocation testing is warranted, but because of the risk of severe bronchoconstriction this must be conducted by specialised staff with facilitiesfor emergency resuscitation.

Table 5

Guidelines for use of analgesics in asthmatic patients

View this table:

We found that a significant proportion of asthmatic patients are sensitive to aspirin. Many maybe unaware of their sensitivity because either they have never taken aspirin or they developed aspirin induced asthma in adulthood after years of apparent tolerance. Since aspirin and NSAIDs are often self prescribed, patients diagnosed with asthma should be alerted to the possibility of aspirin induced asthma by their healthcare professional. Our data justifies the need to include simple,standardised warnings on packs of aspirin and NSAIDs, alerting asthmatic patients to the potentialrisks.

What is already known on this topic

Aspirin induced asthma is a distinct syndrome that is potentially life threatening

The prevalence and cross reactivity to other analgesics has been difficult to assess due to differences in trial methods

Asthmatic patients sensitive to aspirin are usually cross reactive to NSAIDs but seldom react to paracetamol

What this study adds

Aspirin induced asthma is more prevalent than previously suggested

Less than 2% of asthmatic patients are sensitive to both aspirin and paracetamol; reactions to paracetamol tend to be less severe


  • Contributors All authors conceived and initiated this review. CJ performed the analysis and drafted the paper; she will act as guarantor for the paper. JC interpreted the results and helped with writing. LH helped with the research, interpretation, and writing.

  • Funding CJ and LH are clinicians working partly in public hospital and private practice and in a research institute that receives funding from government, bequests, and many industry sources. Neither is directly funded by any manufacturer of non-steroidal anti-inflammatory drugs or paracetamol

  • Competing interests CJ has received payment from GlaxoSmithKline Consumer Healthcare Australia, the manufacturer of Panadol (paracetamol), for speaking at a conference. JC serves as a consultant on the Global Analgesics Advisory Board, which is funded by GlaxoSmithKline Consumer Healthcare

  • Ethical approval: None required