Published ahead of print on January 7, 2004, doi:10.1164/rccm.200304-596OC
© 2004 American Thoracic Society
Prospective Study of Acetaminophen Use and Newly Diagnosed Asthma among WomenDivision of General Medicine, Department of Medicine, College of Physicians & Surgeons, and Division of Epidemiology, Mailman School of Public Health, Columbia University, New York, New York; Channing Laboratory, Department of Medicine, Brigham & Women's Hospital; Department of Emergency Medicine, Massachusetts General Hospital, Harvard Medical School; Departments of Epidemiology and Environmental Health, Harvard School of Public Health, Boston, Massachusetts Correspondence and requests for reprints should be addressed to R. Graham Barr, M.D., Dr.P.H., Division of General Medicine, PH-9E 105 Columbia Presbyterian Medical Center, 622 West 168th Street, New York, NY 10032. E-mail: rgb9{at}columbia.edu
Acetaminophen decreases glutathione levels in the lung, which may predispose to oxidative injury and bronchospasm. Acetaminophen use has been associated with asthma in cross-sectional studies and a birth cohort. We hypothesized that acetaminophen use would be associated with newly diagnosed adult-onset asthma in the Nurses' Health Study, a prospective cohort study of 121,700 women. Participants were first asked about frequency of acetaminophen use in 1990. Cases with asthma were defined as those with a new physician diagnosis of asthma between 1990 and 1996 plus reiteration of the diagnosis and controller medication use. Proportional hazard models included age, race, socioeconomic status, body mass index, smoking, other analgesic use, and postmenopausal hormone use. During 352,719 person-years of follow-up, 346 participants reported a new physician diagnosis of asthma meeting diagnostic criteria. Increasing frequency of acetaminophen use was positively associated with newly diagnosed asthma (p for trend = 0.006). The multivariate rate ratio for asthma for participants who received acetaminophen for more than 14 days per month was 1.63 (95% confidence interval, 1.11–2.39) compared with nonusers. It would be premature to recommend acetaminophen avoidance for patients with asthma, but further research on pulmonary responses to acetaminophen is necessary to confirm or refute these findings and to identify subgroups whose asthma may be modified by acetaminophen.
Key Words: analgesics • paracetamol • aspirin • analgesic drugs, nonsteriodal Asthma prevalence has increased dramatically since the 1970s and currently affects 5 to 8% of the population of the United States (1). Concurrent increases in asthma-related hospitalization and mortality suggest that the change in asthma prevalence did not result from greater diagnosis and detection alone (2), although asthma-related hospitalization and mortality appear to have declined since 1995 with the more widespread use of inhaled corticosteroids (3). Various hypotheses have been proposed to explain the rise in asthma prevalence, including those relating to changes in early life antigen exposure (4) and to the obesity epidemic (5, 6). The rise in the prevalence and severity of asthma in the United States, however, also coincided with a large increase in the use of acetaminophen in the 1970s and 1980s. The switch from aspirin to acetaminophen was particularly pronounced among children following reports that linked aspirin to Reyes Syndrome in the early 1980s (7). This substitution of acetaminophen for aspirin was not evaluated in randomized trials. By contrast, ibuprofen was recently compared with acetaminophen for pediatric febrile illness in a large randomized, double-blind, clinical trial (8). Among the subgroup of 1,879 children with asthma, asthma-related outpatient visits were significantly lower in the ibuprofen arm, and asthma hospitalizations were nonsignificantly reduced compared with the acetaminophen arm (9). The trial did not include a placebo arm, therefore it is uncertain whether ibuprofen decreased or acetaminophen increased asthma morbidity. Alternatively, the finding may have been due to chance. An increase in asthma risk related to acetaminophen use was also suggested by a population-based case–control study (10). The study was limited, however, by the case–control design in which the diagnosis of asthma preceded ascertainment of acetaminophen use. We therefore analyzed data from a prospective study to examine if acetaminophen use was associated with a new physician diagnosis of asthma among participants not previously diagnosed with asthma. Some of the results have been previously reported in the form of an abstract (11).
In 1976, the Nurses' Health Study enrolled 121,700 married female registered nurses, aged 30 to 55 years, who resided in one of 11 U.S. states and who completed a mailed questionnaire (12). We updated the information with biennial follow-up questionnaires and inquired about a physician diagnosis of asthma from 1988 onward. Follow-up of the original cohort was greater than 90%.
Ascertainment of Analgesic Use
Definitions of Asthma Cases with asthma were defined as those with a report of a physician diagnosis of asthma on both the regular questionnaire (1990–1996) and the supplementary questionnaire (1998) plus use of an asthma-controller medication (inhaled or systemic corticosteriod, theophylline, leukotreine modifier, or cromolyn) in the preceding 12 months. Cases reporting comorbid pulmonary disease (e.g., sarcoidosis) were excluded. This asthma definition was validated in a parallel study (5) (see online supplement for details).
Study Sample
Statistical Analysis Incidence rates were calculated as the number of new cases divided by the person-time in each category of acetaminophen use, and rate ratios were defined as the incidence rate of asthma in each category of acetaminophen use divided by the corresponding incidence rate among nonusers. Age-adjusted rates were calculated, and Mantel–Haenzel ratios (13) and 95% confidence intervals were estimated (14). Cox proportional hazards models (15) were used to adjust the rate ratio for newly diagnosed asthma for covariates (listed in Table 2). Specification of multivariate models is described in the online supplement. Secondary analyses included (1) restriction to participants who reported consistent analgesic use, defined as a change in reported analgesic use of no more than two categories of use between 1990 and 1992; (2) inclusion in multivariate models of indices of depressive symptomatology and role limitation due to emotional problems from the Medical Outcomes Study Short-Form 36 Health Survey (16, 17), which were available only for 1992; (3) use of date of first asthma symptoms rather than the date of diagnosis as the incidence date. All p values were two tailed, with p values less than 0.05 being considered statistically significant. Analyses were performed using SAS 6.12 (SAS Institute, Cary, NC).
During 352,719 person-years of follow-up from 1990 to 1996, 346 participants reported a new physician diagnosis of asthma that fulfilled diagnostic criteria. Participants who reported no acetaminophen use accounted for 39% of follow-up person-time. The distribution of follow-up person-time by average acetaminophen use was: 1 to 4 days per month, 28% of person-time; 5 to 14 days per month, 9%; 15 to 21 days per month, 3%; and 22 days per month or more, 5%. Information on acetaminophen use was missing for the remaining 16% of person-time. Acetaminophen use remained fairly stable over time. Sixty-two percent of participants who reported no acetaminophen use in 1990 also reported no acetaminophen use in 1992. Similarly, 73% of participants who reported acetaminophen use in 1990 reported usage in 1992. Overall, less than 3% of participants reported a change in acetaminophen use of two or more categories between 1990 and 1992. Women who used acetaminophen had generally similar demographics and health habits as women who did not use acetaminophen (Table 1). Women who frequently used acetaminophen were more likely to frequently use aspirin and other NSAIDs; conversely, women who did not use acetaminophen were less likely to use other analgesics (Table 1).
The age-adjusted rate ratio for new physician-diagnosed asthma rose with increasing frequency of acetaminophen use (Table 2, p for trend < 0.001). Adjustment for use of other analgesics attenuated the association only slightly, as did multivariate adjustment for age, time period of diagnosis, other analgesic use, race/ethnicity, socioeconomic status, region, smoking, secondhand smoke exposure, body mass index, postmenopausal hormone use, and type of menopause (Table 2). The multivariate rate ratio for newly diagnosed asthma for participants who received acetaminophen for more than 14 days per month was 1.63 (95% confidence interval, 1.11–2.39) compared with nonusers of acetaminophen.
Because analgesic use was assessed at only two time intervals and some participants altered their acetaminophen use over time, we restricted the analysis to consistent acetaminophen users, excluding the 3% of participants who reported a change in acetaminophen use of more than two categories between 1990 and 1992. The results of this better-classified analysis showed a monotonic relationship between frequency of acetaminophen use and asthma (Figure 1, p for trend = 0.008). Further restriction to participants who reported exactly the same acetaminophen use in 1990 and 1992 produced a similar relationship. Defining exposure by 1990 values of analgesic use (i.e., analyzing the data without updating the exposure) yielded similar results (p for trend = 0.002) as did defining exposure by 1992 values of analgesic use (p for trend = 0.04).
The positive association of acetaminophen use and asthma did not differ significantly between participants who did and did not receive aspirin (p for interaction = 0.52). There was therefore no statistical evidence for a difference in the association of acetaminophen and asthma by subgroups of aspirin use. The rate ratio for asthma increased with increasing acetaminophen use in the subgroup that received aspirin, although the smaller sample size increased the variability in the estimates (Table 3). A similar pattern was observed in the subgroup that did not receive aspirin, although this did not attain statistical significance in the multivariate analysis.
Further analyses restricting the asthma case definition to never smokers excluding participants with heart disease and including those without physician visits produced qualitatively similar results, as did adjustment for depression, role limitation due to emotional problems, and nutritional factors such as vitamin E intake and defining the index date by onset of asthma symptoms rather than by asthma diagnosis (data not shown). Many participants also received aspirin and other NSAIDs, as described in Table 1. Frequency of use of other NSAIDs was significantly associated with newly diagnosed asthma in analgesic-adjusted analyses but was not significantly associated with newly diagnosed asthma in multivariate analyses (p for trend = 0.12; Table 4). Restricting the analysis to consistent users of other NSAIDs produced a similar result (p for trend = 0.10).
Frequency of aspirin use was not associated with newly diagnosed asthma in analgesic-adjusted analyses but was inversely associated with newly diagnosed asthma in multivariate analyses (p for trend = 0.03; Table 5). The exclusion of participants who reported a change in aspirin use of more than two categories produced a similar result (p for trend = 0.01), as did the exclusion of participants with heart disease (p for trend = 0.04).
During 352,719 person-years of follow-up of participants in the Nurses' Health Study, greater use of acetaminophen was prospectively associated with increased rate of new physician diagnosis of adult-onset asthma. Our findings confirm and extend the findings of prior cross-sectional studies of asthma and acetaminophen use. Across countries in Europe, per capita consumption of acetaminophen was ecologically associated with the prevalence of wheeze, diagnosed asthma, and bronchial hyperresponsiveness (18). In addition to the ecologic findings, a population-based, case–control study from the United Kingdom showed a dose-dependent relationship between acetaminophen use and asthma, with a multivariate odds ratio for asthma of 2.4 (95% confidence interval, 1.2–4.6) comparing daily users with never users (10). The association was much stronger for severe asthma (odds ratio 8.2; 95% confidence interval, 2.8–23). Aspirin use was equally common among cases and control subjects (odds ratio 1.05; 95% confidence interval, 0.47–2.35). Although aspirin avoidance was slightly more common among cases than among control subjects, the magnitude of the difference in that study was not large enough to explain the association of acetaminophen and asthma. Our findings are also consistent with a recently published study of acetaminophen use in pregnancy, which showed that acetaminophen use in late pregnancy was associated with an increased risk of wheeze among offspring (19). Acetaminophen-related bronchospasm has been reported for at least 35 years in a subset of patients with asthma (20, 21). Acetaminophen provokes bronchospasm in up to 32% of patients with stable, aspirin-sensitive asthma (22, 23). Reactions generally are milder than seen after aspirin challenge and occur with a high, but clinically relevant, dose of acetaminophen (e.g., 1,000 mg). Acetaminophen-related bronchospasm also has been demonstrated in some patients with no history of aspirin-sensitive asthma (24). The mechanism for this phenomenon is unclear, but may involve glutathione, an antioxidant (in its reduced state) that mitigates lung oxidative stress (25, 26). Acetaminophen decreases levels of glutathione in the liver, kidneys, and lungs (27, 28). These decreases are dose dependant: overdose levels of acetaminophen are cytotoxic to pneumocytes and cause acute lung injury (29), whereas nontoxic, therapeutic doses produce smaller, but significant, reductions in glutathione levels in type II pneumocytes and alveolar macrophages (30). Oxidative stress in asthma occurs from the production of reactive oxygen species in the lung by inflammatory cells (31). Reactive oxygen species cause contraction of airway smooth muscle and release of leukotrienes and other secondary inflammatory mediators, leading to bronchial hyperresponsiveness and bronchoconstriction (31). The importance of the glutathione pathway in asthma is reinforced by the finding that polymorphisms in glutathione-S-tranferase are associated with increased susceptibility to pediatric asthma and with slowed lung function growth in childhood (32, 33). Although we were interested primarily in the association of acetaminophen use and asthma, we also evaluated use of other over-the-counter analgesics. An unanticipated finding was that aspirin use was inversely associated with newly diagnosed, adult-onset asthma. This result may be a chance finding because, in contrast to the findings for acetaminophen, the association of aspirin use and asthma was not significant in analgesic-adjusted analyses and remained close to the standard threshold for statistical significance in multivariate analyses. An alternative explanation for the aspirin finding is "depletion of susceptibles" before the start of the follow-up period. Asthma is a heterogeneous disease; aspirin provokes bronchospasm in the subgroup of patients with aspirin-sensitive asthma (34) but improves bronchospasm in other patients with asthma (35–37). Given secular trends in analgesic consumption in the United States, the study participants (mean age, 57 years) were likely to have been exposed to aspirin as children or young adults. Those prone to aspirin-sensitive asthma were likely to have been diagnosed with asthma in childhood or as young adults and consequently excluded from this analysis, leaving only those for whom aspirin had no effect or an ameliorative effect. Given that acetaminophen was not widely prescribed to children and young adults in the United States before the 1970s, no similar selective pressure for acetaminophen occurred in this older cohort. A third possible explanation, given the observational design of the study, is a form of confounding by indication (38). The prospective design, however, ensured that, for confounding by indication to occur, persons with asthma symptoms or at a high risk of asthma must have preferentially avoided aspirin and used acetaminophen—before they were diagnosed with asthma. We evaluated for the presence of such confounding by indication by repeating analyses on the basis of date of onset of asthma symptom rather than date of diagnosis. We also inquired about remission and relapse of early-onset asthma and excluded participants who reported a relapse, rather than a new diagnosis, of asthma. Although some participants may have forgotten and therefore not reported childhood asthma that had remitted, we believe it unlikely that many such participants avoided aspirin in adulthood—particularly in this cohort whose participants' first 2 decades of life (the 1930s and 1940s, on average) preceded most, although not all, of the literature on aspirin-sensitive asthma. Finally, a large randomized, double-blind, clinical trial that compared acetaminophen and ibuprofen showed a higher rate of asthma morbidity among children randomized to acetaminophen (9). Because the trial did not include a placebo group, the direction of causal effects could not be determined; the trial does, however, suggest that observed differences in asthma may not be due to confounding by indication. Other potential known confounders such as body mass index were measured with validated instruments. Participants with depression, anxiety, feelings of limitation due to emotional problems, or headaches/migraine may be more likely to receive analgesics and be diagnosed with asthma. Psychosocial measures were available from 1992, and adjustment for these measures did not appreciably alter the findings. Bias from unknown and unmeasured confounders is always a possibility, given the observational nature of the study. Frequency of analgesic use was assessed at two time intervals. More frequent measures of analgesic use and more quantitative information, such as doses of analgesics would have been desirable. Most participants in the study, however, were fairly consistent in their analgesic use over 2 years. When analyses were restricted to participants who reported consistent analgesic use, the association of acetaminophen use and asthma became monotonic. The Nurses' Health Study sample is not representative of the general U.S. female population and not all participants in the Nurses' Health Study provided information of analgesic use. Given the prospective design of this study, the nonrepresentative nature of this sample should not bias the study's findings but may affect its generalizability. We do not believe, however, that the putative mechanism of acetaminophen and asthma differs by race/ethnicity or socioeconomic class; therefore, we expect findings of this study to be generalizable. Furthermore, the above-mentioned case–control study (10) and randomized trial (9) were population based and practitioner based, respectively. In summary, we found that acetaminophen use was associated with an increased rate of newly diagnosed, adult-onset asthma. This finding suggests that the population-level increases in asthma in the United States might be attributable, in part, to historical trends in analgesic use. Asthma is a heterogeneous disorder. Clearly, many patients with asthma tolerate acetaminophen without problems, and a small subgroup of patients with asthma react severely to aspirin and, less frequently, with other NSAIDs. The U.S. National Asthma Education and Prevention Program recommends counseling patients with severe persistent asthma or a history of aspirin sensitivity about aspirin avoidance (39). It would be premature, on the basis of the current report, to recommend the avoidance of acetaminophen for all adults with asthma. Children were not included in our study. Further epidemiologic and biologic research on the pulmonary responses to over-the-counter analgesics is indicated to better understand these phenomena and to examine if patient subgroups exist that might benefit from a more deliberate selection of analgesic medications for regular use.
The authors thank Karen Corsano and Gary Chase for invaluable assistance with the implementation of the study and Steven Shea, M.D., for helpful comments on the manuscript.
Supported by a Robert Wood Johnson Generalist Physician Faculty Scholar Award to R.G.B., and grants HL-07427, HL-63841, and CA-87969 from the National Institutes of Health. This article has an online supplement, which is accessible from this issue's table of contents online at www.atsjournals.org Conflict of Interest Statement: R.G.B. has no declared conflict of interest; C.C.W. has no declared conflict of interest; G.C.C. has no declared conflict of interest; S.C.S. received an unrestricted educational grant from Eisai, Inc. and Janssen Pharmaceutical, Inc. in 2000 to study Gastroesophageal Reflux Disease (GERD) and asthma; M.J.S. has no declared conflict of interest; J.S. has no declared conflict of interest; F.E.S. has no declared conflict of interest; C.A.C. has no declared conflict of interest. Received in original form April 30, 2003; accepted in final form December 31, 2003
This article has been cited by other articles:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
