Acetaminophen Use and the Risk of Asthma in Children and Adults: A Systematic Review and Metaanalysis FREE TO VIEW

Acetaminophen, which is also known as paracetamol, is sold under the brand name of Panadol (GlaxoSmithKline; Middlesex, UK) in the United Kingdom, and Tylenol (McNeil Consumer Healthcare; Ft. Washington, PA) in the United States and Canada. Acetaminophen is one of the most commonly prescribed analgesic agents and is possibly the most prescribed antipyretic in the world. It is estimated that the annual sales of acetaminophen products in the United States alone is roughly $1 billion.¹ The estimated annual sales of combined generic and brand-name acetaminophen products in both Canada and the United States far exceed this figure. In the United Kingdom, 80 million acetaminophen tablets were sold between 2001 and 2002.²

In the past few years, there has been an increase in the global prevalence of asthma.³ In light of the increase in the worldwide use of acetaminophen and the ability of the drug to lower the levels of glutathione, an endogenous antioxidant found in lung tissue, it has been postulated that acetaminophen may play a role in the pathogenesis of the disease.⁴ Numerous case reports,⁵ oral challenge tests,⁶ and epidemiologic studies⁷ have attempted to address this possible harmful association. However, many unanswered questions remain, most notably the magnitude of this association among children and adults. Another important question is the risk of asthma with respect to the timing of acetaminophen intake, especially in investigating confounding by indication, which has been suggested as an alternative explanation for the observed association.⁸ We sought to systematically search the literature in order to quantify the risk of asthma and wheezing among acetaminophen users in both the pediatric and the adult populations. We also examined the effect of prenatal acetaminophen use and the risk of asthma in young children.

We conducted a systematic review and metaanalysis based on Quality of Reporting of Metaanalyses⁹ criteria for reporting the results of systematic reviews of randomized controlled trials (RCTs) and the Metaanalysis of Observational Studies in Epidemiology¹⁰ criteria for reporting results of systematic reviews of observational studies.

We searched for articles published in the English language using major bibliographic databases that index health research. The following databases were searched up to the end of October 2008: MEDLINE (1966); EMBASE (1980); Cochrane Central Register of Controlled Trials (1991); Database of Abstracts of Reviews of Effects (1991); American College of Physicians Journal Club (1991); International Pharmaceutical Index (1970); BIOSIS Previews (1969) [all searched through OvidSP]; and Web of Science (1961). Filters were used in MEDLINE and EMBASE to capture RCTs and observational studies. The initial search strategy was developed from the MeSH terms “asthma and acetaminophen” or “acetaminophen” in MEDLINE. Titles were reviewed for relevance from this search, and subject headings and abstracts were then examined. Appropriate MeSH and key words were added to the search strategy.

The scope notes in MEDLINE and EMBASE also were examined to ensure that the correct MeSH terms were being used. Accordingly, other subject headings were included based on previous indexing and the inclusion of key words based on synonyms used in the scope notes. Consequently, the following broader MeSH terms were used for asthma: “respiratory sounds”; “allergens”; “bronchial diseases”; “respiratory tract diseases”; “bronchitis”; “hypersensitivity”; “respiratory hypersensitivity”; “hypersensitivity, immediate”; “hypersensitivity, delayed”; and “bronchoconstriction.” A number of related key words also were included, for example, “wheez,” “cough,” “whistl$ adj10 chest, (respiratory adj3 hypersen$ or inflam$ or allerg$).” More than 100 terms for acetaminophen were included as key words, such as “acet#minophen$” and “acetaminophen,” and various brand names, including “Tylenol,” “Panadol,” and “Tempra” (Mead Johnson Nutrition; Evansville, IN).

This MEDLINE search strategy was then adapted for the other databases. Proceedings and conference abstracts were searched through the databases PapersFirst (1993) and ProceedingsFirst (1993). Authors' names and the year of published work from key articles were entered into the cited reference search in the Web of Science. We also checked the references of retrieved articles for any potentially missed articles.

We considered all clinical studies that compared the risk of asthma or wheezing with acetaminophen and placebo use, in the case of RCTs, or nonacetaminophen users, in the case of observational studies. Observational studies were included if they clearly defined asthma or wheezing as either primary or secondary outcomes; clearly defined acetaminophen use as either primary or secondary exposure, adjusted for potential confounders, and presented as odds ratios (ORs) and their corresponding CIs; or provided enough data to compute these parameters. In cases in which a study was published in different phases or where data from a trial were duplicated in more than one study, we only included the most recent study.

Each study was scored for its quality. For observational studies, we created a quality score based on previously published scales for observational studies.¹¹ We created three separate scales for cross-sectional studies, cohort studies, and case-control studies. Cross-sectional studies were assessed based on outcome definition, exposure definition, and risk adjustment. Case-control studies were assessed based on case definition, control selection, exposure definition, and risk adjustment. Cohort studies were assessed based on outcome definition, cohort entry definition, exposure definition, and risk adjustment. All categories within each scale were scored from 0 to 2. We have used a similar approach in previous systematic reviews.¹² Studies were scored by two blinded authors, and discrepancies were resolved by consensus.

We created a data abstraction sheet and recorded the following study characteristics: authors' names; study design; sample size; study population type; mean age or age range; confounders adjusted for; adjusted OR; and 95% CI. Data were extracted for two separate outcomes of asthma and wheezing, and for each study we extracted data on the current or regular use of acetaminophen. If only one level of exposure was presented, it was assumed to be regular use or current use. If the authors had not defined current or regular use and more than one category was reported, the median category was taken as “regular use.” For studies that provided several levels of acetaminophen intake, the highest level was considered the “high intake.” For studies that looked at prenatal exposure and reported ORs for only 0 to 20 weeks or 20 to 32 weeks, we combined the ORs using a fixed-effects model to estimate the pooled OR for the prenatal exposure period.

We performed four main analyses that looked at the risk of asthma and wheezing among adults and children. We also looked at the effect of prenatal exposure to acetaminophen and the subsequent risk of asthma and wheezing in children.

For all analyses, we weighted the study-specific, adjusted log ORs for case-control, cohort, or cross-sectional studies by the inverse of their variances. Relative risks were considered an approximation of the ORs. A random-effects model was used to estimate the pooled adjusted OR. We assessed heterogeneity by computing the Q statistic and its bootstrap version with 1,000 replications.¹³ We also computed the I² statistic, which is the percentage of total variation in the studies that is due to heterogeneity. The I² is calculated simply as D − Q/Q ×100, where Q is the value for the Q statistic and D is the degrees of freedom. Finally, we used metaregression to further identify the source of heterogeneity in our analysis. To do so, we regressed log ORs for each study on study score and study weights. Because the number of studies in each analysis was small, we believed that a funnel plot was not the optimal approach to assess publication bias; therefore, we used the regression of Sterne et al¹⁴ to assess the presence of publication bias. All analyses were performed using a statistical software package (STATA, version 10.0 for Windows; STATA Corporation; College Station, TX).

Twenty-four studies^7,15–37 met our inclusion criteria. After carefully reviewing the articles, four studies^15–18 were found to be duplicates and part of the larger International Study of Asthma and Allergies in Childhood (ISAAC).²⁶ One study¹⁹ was excluded because it was part of the same study published at a later date.³³ Thirteen cross- sectional studies,^{7,21,22,25–32,36,37} 4 cohort studies,^24,33–35 and 2 case-control studies^20,23 comprising 425,140 subjects were included in the review (Fig 1, Table 1). Nine studies^{7,20–22,24,26–28,33} looked at the risk of asthma in both children and adults (Fig 2, Table 2). Six studies^{25,27,28,30–32} looked at wheezing in children, and five studies^33–37 compared prenatal exposure to acetaminophen and wheezing in children. The OR for asthma among users of acetaminophen (adult and children) was 1.63 (95% CI, 1.46 to 1.77) [Table 2]. The risk of asthma in children among users of acetaminophen in the year prior to asthma diagnosis and within the first year of life was elevated by similar magnitude (OR: 1.60 [95% CI, 1.48 to 1.74] and 1.47 [95% CI, 1.36 to 1.56], respectively). Only one study²⁶ reported high-dose acetaminophen use in children, and its result was consistent with an increase in the risk of asthma (OR, 3.23; 95% CI, 2.9 to 3.6). The risk of asthma with acetaminophen use in adults was 1.74 (95% CI, 1.36 to 2.28) [Fig 3]. Six studies^{25,27,28,30–32} assessed acetaminophen use and wheezing in children in the past year, with a pooled OR of 1.97 (95% CI, 1.51 to 2.56) [Fig 4]. There was an increase in the risk of asthma and wheezing with prenatal acetaminophen use (OR: 1.28 [95% CI, 1.13 to 1.39] and 1.50 [95% CI, 1.10 to 2.05], respectively). The Egger regression¹⁴ did not indicate publication bias for the analysis of asthma in adults (p = 0.08) and wheezing (p = 0.90) in children. The results of the metaregression analysis did not indicate heterogeneity due to the quality score (p = 0.07).

The results of our systematic review indicate an increase in the risk of asthma with acetaminophen use in both adults and children. We also found an increase in the risk of asthma in children with a history of prenatal exposure to acetaminophen. The strength of our review is mainly in the large sample size, the multinational nature of the study participants, and the consistency of the effect in both children and adults.

Confounding bias has been deemed to be one of the reasons for the positive association with acetaminophen and asthma observed in observational studies. One form of confounding by indication⁷ suggests that children with more severe forms of asthma are more likely to be receiving acetaminophen for the treatment of viral infections or other respiratory infections that may be due to asthma or may precede an asthma diagnosis, making the association appear to be causal. Another form of confounding by indication, also referred to as “reverse causality bias,”³⁸ arises when a drug is used to treat early manifestations of a disease. For example, if acetaminophen is used to treat a fever that may precede asthma symptoms, it may appear that acetaminophen increases the risk of asthma. Although we cannot disregard the possibility of confounding as a potential explanation for our results, it is unlikely that bias alone may be the sole explanation. As Beasley et al²⁶ have suggested, due to the multinational nature of the studies and the differences in medical practice, it is highly probable that acetaminophen was used to treat fever due to a wide array of diseases unrelated to respiratory infections, including malaria, otitis media, and measles, making confounding by indication less likely. The argument that the results may be explained by reverse causality bias also is questionable because fever is a common phenomenon among infants and is likely to be treated by acetaminophen in both case patients and control subjects, resulting in a null association rather than in a positive association. One approach to better addressing reverse causality bias is to allow a longer time window between acetaminophen use and asthma diagnosis. This scenario is apparent in prenatal exposure to acetaminophen, in which the children of women who received acetaminophen during pregnancy were more likely to receive an asthma diagnosis both in early and in late childhood.^34–36

Several mechanisms have been postulated that may explain the possible risk of asthma with acetaminophen use. The main mechanism involves glutathione, which is an antioxidant found in lung tissue.³⁹ Acetaminophen has shown to lower serum glutathione levels in healthy volunteers.⁴⁰ Glutathione is a potent scavenger of free radicals,⁴¹ which in turn are responsible for lung injury and the initiation of an inflammatory response that leads to bronchoconstriction and asthmatic symptoms. The studies⁴⁰ that have shown inhibition of glutathione by acetaminophen have used relatively high doses of acetaminophen (4 g/d), which may not explain the risk of asthma observed in ISAAC,²⁶ where lower doses of the drug were used. A second possible mechanism is thought to be the lack of inhibition by acetaminophen of the enzyme cyclooxygenase,⁴² which is the key enzyme involved in the production of prostaglandins, which play a major role in the inflammatory cascade in asthma patients. Finally, an emerging hypothesis^43,44 involves the possible antigenic effect of acetaminophen and the subsequent rise in IgE and histamine levels with exposure to acetaminophen.

Our results have important clinical implications because acetaminophen is widely used both as an antipyretic and an analgesic agent in children and adults. Ibuprofen possibly is the second most used antipyretic agent. Unlike aspirin, ibuprofen has not been linked to Reye syndrome. Although the risk of the development of asthma with the use of ibuprofen has not been addressed in an RCT, when compared to acetaminophen, ibuprofen has been shown to have a lower risk of asthma (OR, 0.56; 95% CI, 0.34 to 0.95).⁴⁵ However, it is difficult to determine the safety of ibuprofen over acetaminophen with respect to asthma based on one study. Although RCTs may be the ideal study design to answer this question, due to ethical concerns such studies may not be feasible in researching drug-related adverse events. A better alternative may be a well-designed cohort study with adequate information on acetaminophen dose and potentially confounding variables, and a long follow-up period.

Our systematic review is subject to several limitations. First, all the studies included in the review are observational studies. Although most controlled for potential confounding variables, there is still a possibility that bias may have affected the results of this study. Ascertainment of exposure was also heterogeneous among the different studies, which may have affected our results. Finally, ascertainment of asthma diagnosis in most of the studies was through self-reporting. We cannot exclude the possibility of misclassification of asthma with other respiratory conditions, including wheezing.

In conclusion, the results of our review are consistent with an increase in the risk of asthma in both children and adults who have been exposed to acetaminophen. Future studies are needed to confirm these findings.

Author contributions: Dr. Etminan contributed to all aspects of this study in addition to providing administrative, technical, and material support and study supervision. Dr. Etminan also has full access to all of the data and has taken responsibility for the integrity and accuracy of the data. Dr. Sadatsafavi contributed to the study concept and design, analysis and interpretation of the data, statistical analysis, and critical revision of the manuscript. Dr. Jafari contributed to the acquisition of data and critical revision of the manuscript. Ms. Doyle-Waters contributed to critical revision of the manuscript. Dr. Aminzadeh contributed to the study concept and design, acquisition of data, and critical revision of the manuscript. Dr. FitzGerald contributed to drafting the manuscript and its critical revision.

Financial/nonfinancial disclosures: The authors have reported to the ACCP that no significant conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.