Description of the condition.

Nonsteroidal anti-inflammatory drug (NSAID)-exacerbated respiratory disease (NERD) was first described 50 years ago by Samter and Beers and was previously known as aspirin-induced asthma or aspirin-exacerbated respiratory disease (AERD) [1]. NERD is a chronic eosinophilic inflammation of the respiratory tract accompanied by nasal polyps, chronic rhinosinusitis and/or asthma, in which the symptoms are typically exacerbated by NSAIDs, including aspirin (ASA) [1, 2]. NERD requires follow-up by several specialties, including pulmonology to manage difficult-to-control asthma, allergology for the management of hypersensitivity to NSAIDs, chronic eosinophilic inflammation, and otolaryngology due to the recurrence of nasal polyps and requirement of surgery [1].

The prevalence of NERD varies from 1.8–44%, depending on the study population and the diagnostic criteria used [1]. The Global Allergy and Asthma European Network GA2LEN reported that 1.94% of the population presents dyspnea associated with NSAID consumption, with an increase in the risk of asthma 4 times greater in patients with NERD [4]. The risk of NERD increases in parallel with the severity of respiratory disease, and these patients have higher hospitalization rates due to asthma (NERD 11.8% vs without NERD 2.4%) [4]. In patients with hypersensitivity to NSAIDs confirmed by a provocation test, the prevalence of asthma increases up to 21% [1–4]. In the univariate analysis of GALEN, an increased risk of asthma [OR 5.50 (4.84–6.26)] and chronic rhinosinusitis [OR 4.28 (3.78–4.84)] was reported in this population [4]. Patients with NERD have twice the risk of having uncontrolled asthma, 60% more asthma exacerbations, 80% more emergency consultations, and 40% more hospitalizations. Additionally, they require more asthma medications and have a poorer quality of life than patients without NERD [5–7].

Among the risk factors for developing NERD, a family history of the disease, the presence of nasal polyps associated with chronic rhinosinusitis and/or asthma, and atopy stand out, alongside a slight predisposition of female patients compared to the male population [1, 8–10]. The disease is usually diagnosed in the 3^rd - 4^th decade of life, and its natural history involves chronic rhinitis as the first manifestation, progressing to chronic rhinosinusitis, nasal polyps, and anosmia. During the latter period, asthma appears to be triggered [8] and often occurs before acquiring hypersensitivity to NSAIDs. However, there are cases in which hypersensitivity to NSAIDs occurs before the onset of chronic airway disease [1]. Despite NSAID avoidance, patients continue to have asthma exacerbations, loss of smell, and the need for multiple sinus surgeries [11].

After the intake of NSAIDs, symptoms appear within 30–180 minutes, the onset and severity of which are associated with the dose administered. Most patients develop symptoms with 60 mg of acetylsalicylic acid (ASA), but this range varies from 10–300 mg [1, 12, 13]. Manifestations are characterized by high levels of respiratory symptoms, such as nasal congestion and rhinorrhea, and may progress to wheezing, coughing, and dyspnea. In patients with uncontrolled asthma, symptoms appear more rapidly and severely and could potentially lead to fatal outcomes [14]. Urticaria and gastrointestinal (GI) symptoms are also common [1, 9]. Less frequently, patients manifest symptoms associated with alcohol consumption, with eosinophilia sometimes observed in the blood work [1].

Clinical history is key to making a diagnosis of NERD. The appearance of respiratory symptoms 1–2 hours after the consumption of NSAIDs, in patients with adult-onset asthma and with a history of repeated nasal polyposis are key to identifying patients with NERD. If the patient does not meet all criteria or there is doubt in the diagnosis, NSAID hypersensitivity must be confirmed using an oral provocation test, in which increasing doses of the drug are administered following established protocols. This should be done in a safe environment with adequate staff and equipment to ensure an appropriate response to any reactions, such as conjunctivitis or rhinitis, low respiratory symptoms, bronchospasm, a decline in pulmonary function (a decrease in forced expiratory volume in one second (FEV1) by more than 15%), laryngospasm, cutaneous manifestations and systemic symptoms [1, 13]. Oral provocation tests are contraindicated in patients with previous anaphylactic reactions associated with NSAIDs or ASA, uncontrolled asthma with FEV1 <70% of the predicted value, history of chronic renal failure or gastrointestinal bleeding, an exacerbation of asthma in the previous month, pregnancy and present management with beta-blockers [1].

The mainstay for management is the avoidance of the causative drug and other strong COX-1 inhibitor molecules, such as piroxicam, indomethacin, sulindac, tolmetin, ibuprofen, naproxen, fenoprofen, oxazoprin, mefenamic acid, flurbiprofen, diflunisal, ketoprofen, diclofenac, ketorolac, etodolac, nabumetone, and acetylsalicylic acid; typically, NERD patients tolerate selective COX-2 inhibitors, such as celecoxib and etoricoxib. It is of the utmost importance that patients are educated about their disease, understand the alternative medications that are safe for them, and avoid alcohol consumption, as it can worsen symptoms [1]. Specifically, in the asthmatic population, treatment is performed according to various guidelines developed by The National Heart, Lung, and Blood Institute (NHLBI), The Global Initiative for Asthma (GINA), and The British Thoracic Society (BTS), among others [15]. In the GALEN cohort study, it was reported that patients with NERD consume more medications for asthma control (26.1 vs 5.6%) [4], and approximately 30% of patients with hypersensitivity of NSAIDs require high doses of inhaled corticoids [1]. Additionally, these patients typically benefit from management with leukotriene antagonists to try to reduce the existing overexpression of cysteinyl leukotrienes, and in particular cases, they require management with biologics, such as omalizumab or anti-IL5 molecules [1, 16]. Concerning the management of chronic rhinosinusitis and nasal polyps, patients with NERD are more resistant to the usual pharmacological treatments, such as intranasal steroids, also requiring oral steroids for control of the disease. These patients frequently require multiple surgical reinterventions for the recurrence of nasal polyposis (from 24–80% of patients) approximately every 3 years [1, 2].

Inclusion criteria.

We included patients ≥ 18 years old with a diagnosis of asthma associated with chronic rhinosinusitis and nasal polyps, with a previous history of pulmonary symptoms triggered by ASA or other NSAIDs, or with a positive provocation test to ASA.

Exclusion criteria.

Patients with a history of GI bleeding, bleeding diathesis, uncontrolled arterial hypertension, chronic renal failure, uncontrolled asthma with FEV1 < 70%, autoimmune disorders, malignancy, or pregnancy were excluded.

Patients with a history of other pulmonary diseases, such as cystic fibrosis or primary ciliary dyskinesia, we also excluded.

Primary outcomes.

FEV1 during spirometry

Total daily dosage of systemic steroids

Total daily dosage of inhaled steroids

Acute asthma exacerbations

Secondary outcomes.

Frequency of symptoms assessed with symptom score

Medication need score

Nonfatal adverse events

Electronic searches.

A PubMed, EMBASE, SCOPUS, and EBSCO search was performed in June 2020 using controlled vocabulary (Mesh and Emtree terms), as well as free text with the following search terms: aspirin-exacerbated respiratory disease, aspirin desensitization, acetylsalicylic acid, asthma, and pulmonary disease.

Results were limited to human-based clinical trials, including patients 18 years old or older, written in English or Spanish. References were manually searched for additional relevant studies. We did not filter results by date of publication.

Selection of studies.

Studies were screened by two independent investigators, IE and SS, who analyzed the titles and abstracts for potentially relevant studies. Then, each study was assessed against the preset inclusion and exclusion criteria.

Data extraction and management.

Two investigators, IE and SS, extracted the data from all studies using a standardized data collection form. When disagreements were present, a third party was consulted. For each study, the trial design, participant characteristics, type of interventions, and outcomes were assessed. We used Cochrane Review Manager (RevMan 5.1) software to analyze the data [24].

Assessment of risk of bias in the included studies.

Bias was assessed according to the recommendations outlined in the Cochrane Handbook for Systematic Reviews of Interventions [25]. We analyzed the following items:

Allocation sequence generation

Concealment of allocation

Blinding of participants and investigators

Incomplete outcome data

Selective outcome reporting

We graded each potential source of bias as low risk, high risk, or unclear risk of bias.

Dichotomous data.

We analyzed dichotomous data variables using Mantel-Haenszel odds ratios using a fixed-effects model with 95% confidence intervals. If substantial heterogeneity was found among the studies, a random effects model was chosen. If count data were not reported as the number of events per participant, we transformed the variable into a continuous variable.

Continuous data.

Continuous variables were analyzed as fixed effects mean differences with 95% confidence intervals. If substantial heterogeneity was found among the studies, a random effects model was chosen. Two of the analyzed outcomes used different scales, so the standardized mean difference (SMD) was used. Data were collected using intention-to-treat (ITT) analysis when possible.

Dealing with missing data.

We did not contact the study authors regarding missing data, but we considered this aspect when judging the quality of evidence and in the analysis of results.

Assessment of heterogeneity.

We evaluated the degree of statistical variation using the I² statistic.

Assessment of reporting biases.

For each study, we compared reported outcomes in the methods sections with the published results to check for reporting biases. We did not contact authors for extra information. Furthermore, we considered very few studies to perform funnel plots.

Data synthesis.

We constructed a table to summarize our primary outcomes, called the “Summary of Findings Table,” using GradePro software [26].

Results of the search.

A PubMed, EMBASE, SCOPUS, and EBSCO search was performed in June 2020, which identified a total of 292 articles. After limiting the search using our prespecified filters, 45 records were found. Two additional records were identified through reference searching. After reviewing the selected studies, 18 records were removed because they were duplicates. Twenty-nine articles were screened in total, 15 of which were fully assessed. Finally, 5 studies were included in the qualitative analysis and 4 in the quantitative analysis [27–31] (see the PRISMA 2009 flow diagram in Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA 2009 flow diagram.

https://doi.org/10.1371/journal.pone.0247871.g001

Included studies.

All studies were randomized controlled trials with a parallel design, except for the study by Stevenson et al, which was a crossover trial, which had an appropriate washout period (1 month) to prevent carry-over effects [30]. We decided to exclude such trials from the quantitative analysis due to the trial design and the incompleteness of information on this review´s outcomes. For all studies, there was a combined total of 199 participants. All studies evaluated the efficacy of aspirin desensitization vs placebo in patients with NERD. However, the study by Swierczynska-Krepa et al also performed AD in patients with aspirin-tolerant asthma, which was not taken into account in this systematic review [31]. Therefore, we included a total of 185 participants, of whom 108 patients received AD and 102 received placebo [27–31].

Inclusion criteria were similar across the five studies: participants were ≥ 18 years old with a prior diagnosis of asthma and a history of aspirin hypersensitivity or known aspirin-exacerbated respiratory disease (AERD). To establish aspirin hypersensitivity/AERD, all studies except the one by Fruth et al used an aspirin challenge test, which was considered positive when clinical symptoms (dyspnea, rhinorrhea, sneezing, ocular secretion, skin flushing, cough, etc.) and a decrease in FEV1 between 15 and 25% from baseline were recorded. In the study by Fruth et al., aspirin intolerance was evaluated in vitro by provoking peripheral leukocytes with the posterior measurement of eicosanoids [28]. Exclusion criteria were described in all but the study by Stevenson et al, reporting consistently that patients with FEV1 values < 70% in spirometry, pregnancy or breastfeeding, history of bleeding diathesis/GI bleeding, malignancy or any chronic diseases of the heart, liver, pancreas, urinary or neurologic system were excluded. Standard asthma medications and therapy for rhinosinusitis symptoms were allowed throughout the studies. Only the study by Swierczynska-Krepa et al explicitly reported that leukotriene modifiers, omalizumab, and immunotherapy were not allowed during the study [31]. Across all studies, participants had a mean age of 38.88 years, and 56.22% were female.

All studies used different doses in the desensitization scheme, although the doses of aspirin were escalated throughout the acute desensitization phase (usually 2 days). Three studies by Mortazavi et al, Esmaeilzadeh et al, and Fruth et al used ascending doses until reaching a prespecified maximum dose of 120 mg– 180 mg on day 1 and ascending doses until reaching a maximum dose of 475 mg– 800 mg on day 2 [27–29]. The studies by Stevenson et al and Swierczynska-Krepa et al used a different approach in which escalation continued until the patient developed a reaction or until the maximum dose was reached (650 mg and 624 mg, respectively) [30, 31]. Maintenance therapy with daily aspirin ranged from 100 mg to 800 mg for 3 to 6 months. For more details, see the characteristics of the included studies table (S1 Table).

Allocation.

In all studies, the randomization method was reported. Regarding allocation concealment, 2 studies by Esmaeilzadeh et al and Swierczynska-Krepa et al were judged to have an unclear risk for selection bias. The randomization list was managed by the unblinded study director in the first study, and no relevant information was reported for the latter [27, 31]. Furthermore, in the studies by Mortazavi et al and Esmaeilzadeh et al, patient baseline characteristics were not reported, which raises some concerns [27, 29]. In the latter study, baseline characteristics referred only to baseline outcome measures, which are not necessarily the most relevant confounding variables [29].

Blinding.

All studies were considered to have a low risk of performance and detection bias. They were all double-blinded studies in which participants and personnel were not aware of treatment assignment. Placebo was usually made of lactose or starch and looked identical to aspirin. Additionally, the study medication was managed by the hospital’s pharmacy. In all included studies, participants and investigators remained blinded until the end of the study.

Incomplete outcome data.

An attrition rate of < 5% was considered ideal, being acceptable between 5 and 10%. Taking this into consideration, 3 studies were judged as having a low risk for attrition bias. Although the study by Fruth et al had a withdrawal rate of 55% and no sensitivity analysis was performed, all missing data occurred for documented reasons unrelated to the outcome [28]. On the other hand, the studies by Stevenson et al and Swierczynska-Krepa et al reported withdrawal rates of 34% and 17.4%, respectively, and in this case, missingness of data was related to its true value [30, 31].

Selective reporting.

Complete outcome data were reported according to a prespecified plan in all but one study; only the study by Stevenson et al was judged as having an unclear risk of reporting bias since we could not find the study protocol and pulmonary function (specifically VEF1), which was reportedly measured at every follow-up visit but was not depicted in the results section, raising some concerns [30].

Primary outcomes.

Pulmonary function. Three of the included studies looked at changes in pulmonary functions using VEF1 values (Mortazavi 2017, Esmaelilzadeh 2015, Swierczynska-Krepa, 2014; 86 patients [27, 29, 31]. The study by Swierczynska-Krepa et al did not identify any differences in VEF1 among patients who received AD and those who did not at 6 months [31]. On the other hand, the studies by Mortazavi et al and Esmaelilzadeh et al did report a significant improvement in the experimental group, but in both cases, the increase in FEV1 was seen only after the 6-month follow-up but not earlier [27, 29] (Fig 3). As the study by Swierczynska-Krepa et al reported the results in liters and the other two in % predicted, a standardized mean difference was used. Due to high heterogeneity (I² = 92%) and the small number of studies included, we felt it was not appropriate to pool the data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Forest plot of comparison.

Aspirin desensitization vs placebo, outcome: Pulmonary function (VEF1).

https://doi.org/10.1371/journal.pone.0247871.g003

Total daily dose of systemic steroids. Only the study by Stevenson et al, which was a crossover trial, reported the use of systemic steroids (25 patients) [30]. Although there were available data on the mean daily dosages, it reported how many patients took less, equal, or more prednisone throughout the duration of the study. Of the 25 patients included, 5 were not taking prednisone before the study began. Of those taking prednisone, 5 patients took the same dosage, 9 patients took less, and 6 patients required a higher dose. Of the patients who took a lower dose, an improvement in rhinitis symptoms, asthma symptoms, or both was reported after AD. The remaining patient exhibited a worsening of both symptoms but concomitantly reduced the prednisone dose. Patients who took a higher dose all experienced a worsening of rhinitis and/or asthmatic symptoms during AD. This study did not report changes in prednisone dosage during the placebo phase, which makes it impossible to further analyze this outcome.

Total daily dose of inhaled corticosteroids. Only the study by Swierczynska-Krepa et al reported a daily dosage of inhaled corticosteroids, measured as micrograms (μg) of inhaled budesonide equivalent (15 patients). At 6 months, the daily dosage of inhaled corticosteroids was significantly lower in patients with chronic AD than in the control group, showing a mean difference of 1039.2 μg lower budesonide equivalents, although with a very wide confidence interval (95% CI -1763.4 to -315) (Table 2) [31].

Acute asthma exacerbations. Only the study by Mortazavi et al (38 patients) reported the frequency of acute asthma exacerbations throughout the study, which had a follow-up period of 6 months. The study failed to identify a statistically significant difference between the experimental and control groups, with an OR of 0.40 (95% CI 0.10–1.55) favoring the experimental group (Table 2) [29].

Secondary outcomes.

Symptom score. Three studies reported symptom scores for the experimental and control groups [27–29]. The studies by Mortazavi et al and Esmaelilzadeh et al used the same symptom score, which evaluated nasal (itching, sneezing, nasal discharge, nasal obstruction), eye (itching, redness) and bronchial symptoms (cough, wheezing, difficulty breathing) categorized into mild, moderate and severe, with a maximum score of 27 points (a larger score reflected more severe symptoms) [27, 29]. On the other hand, the study by Fruth et al used a questionnaire that evaluated primary and secondary nasal symptoms (nasal obstruction, postnasal drip, cephalgia, impairment of olfactory function) plus paranasal symptoms (coughing and wheezing), with a maximum score of 20 points [28]. For this review, we were primarily interested in bronchial or paranasal symptoms for each symptom score, since our objective was to evaluate the efficacy of AD in asthma outcomes; however, the results were given as an absolute value without discrimination of the source of the symptoms. All studies found the symptom score to be significantly higher in the control group after 6 months of treatment. As the scores were different, we used the standardized mean difference in the quantitative analysis, favoring the experimental group (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Forest plot of comparison.

Aspirin desensitization vs placebo, outcome: Symptom score.

https://doi.org/10.1371/journal.pone.0247871.g004

Medication score. The studies by Mortazavi et al and Esmaelilzadeh et al reported the medication score, in which local and systemic drugs were evaluated [27, 29]. Local medications included eye drops and nasal sprays (1 point for each). Regarding systemic medications, antihistamines, systemic beta-2 agonists, inhaled steroids, and theophylline were included (2 points for each). If the drug was used at its maximum dose, the score for that drug was multiplied by two. That being said, the study by Mortazavi et al found the medication score to be significantly higher in the control group after 6 months of treatment [29]. In the study by Esmaelilzadeh et al, there was a trend towards better medication scores in the experimental group, but it was not significant [27]. In our quantitative analysis, we observed a pooled mean difference of -1.72 (95% CI -2.13 to -1.32), favoring the experimental group (Fig 5).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Forest plot of comparison.

Aspirin desensitization vs placebo, outcome: Medication score.

https://doi.org/10.1371/journal.pone.0247871.g005

Nonfatal adverse events. All studies reported nonfatal adverse events [27–31], and 12.9% of patients in the experimental group experienced adverse events, with gastrointestinal intolerance being the most common, followed by GI bleeding. The study by Fruth et al reported 0 adverse events in the experimental group [28]. Of the patients in the control group, only 1.9% developed adverse events, corresponding to two patients in the Stevenson et al study. One of those patients presented with uterine bleeding, raising concern of a carryover effect [30]. We included all studies in the quantitative analysis except the one by Stevenson et al, observing a significant difference among groups when the results were pooled, favoring the controls, although with a very wide confidence interval (OR 6.62 (95% CI 1.12–39.25) (Fig 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Forest plot of comparison.

Aspirin desensitization vs placebo, outcome: nonfatal adverse events.

https://doi.org/10.1371/journal.pone.0247871.g006