To cite this article: van den Nieuwenhof L, Schermer T, Bosch Y, Bousquet J, Heijdra Y, Bor H, van den Bosch W, van Weel C. Is physician-diagnosed allergic rhinitis a risk factor for the development of asthma? Allergy 2010; 65: 1049–1055.
Background: There is strong evidence that there is a relationship between allergic rhinitis (AR) and asthma, but it is unclear whether there is a causal relation between AR and asthma. The aim of this study was to assess prospectively whether AR is a risk factor for the diagnosis of asthma in a large primary care population.
Methods: We performed a historic cohort study of life-time morbidity that had been recorded prospectively since 1967 in four general practices. Two groups of subjects were selected: (i) patients with diagnosis of AR, (ii) a control group matched using propensity scores. We assessed the risk of physician-diagnosed asthma in patients with physician-diagnosed AR compared to subjects without a diagnosis of AR (controls).
Results: The study population consisted of 6491 subjects (n = 2081 patients with AR). Average study follow-up was 8.4 years. In patients with AR, the frequency of newly diagnosed asthma was 7.6% (n = 158) compared to 1.6% (n = 70) in controls (P < 0.001). After adjusting the effect of AR on asthma diagnosis for registration time, age, gender, eczema and socioeconomic status, having AR was a statistically significant risk factor for asthma (hazard ratio: 4.86, P < 0.001, 95% confidence interval: 3.50–6.73, controls as reference).
Conclusion: A diagnosis of AR was an independent risk factor for asthma in our primary care study population. Having physician-diagnosed AR increased the risk almost fivefold for a future asthma diagnosis.
- 95% CI
95% confidence interval
Continuous Morbidity Registration
International Classification of Health Problems in Primary Care
There is evidence of a relationship between allergic rhinitis (AR) and asthma, but until now the nature of this link remains a subject of debate (1, 2). Cross-sectional studies have shown that AR and asthma frequently coexist (3). Up to 80% of all patients with asthma have concomitant AR, and over 20% of patients with AR also have asthma (4, 5). A growing number of studies suggest that AR and allergic asthma are manifestations of the same disease entity: ‘one airway, one disease’ (6, 7). Still, not all patients with AR eventually present with asthma and not all patients with asthma have AR. Therefore, prospective studies are required to assess the risk of asthma in patients with AR (8).
Although a causal relationship is suspected, there are only a few longitudinal studies that confirm this widespread notion (9–13). Most of these studies were based on selected populations or consisted of small numbers of study subjects with a high likelihood of selective inclusion. Allergic rhinitis and asthma are frequent disorders in childhood and adulthood and are often encountered in primary care. However, we are not aware of any previous prospective studies in primary care based on physician-diagnosed AR and asthma. Therefore, the aim of this study was to assess prospectively whether a diagnosis of AR is a risk factor for an asthma diagnosis in a large primary care population.
We performed a historic cohort study in a database with complete life-time morbidity data that had been recorded at the time of initial diagnosis in four general practices. We assessed the risk of physician-diagnosed asthma in subjects with physician-diagnosed AR compared to subjects without a diagnosis of AR.
The morbidity data were taken from the Continuous Morbidity Registration (CMR), the academic general practice network of the Radboud University Nijmegen Medical Centre, Department of Primary and Community Care, the Netherlands. The CMR has existed since 1967 (14). All general practitioners (GPs) of these practices enter their diagnoses (morbidity) for each episode of care, uniformly coded and registered. In the structure of the Dutch health care system, everyone is listed with a general practice and can get access to health care through that practice. In terms of sociodemographic characteristics, the population of the CMR practices (approximately 35 500 individuals) is comparable to the Dutch population at large (14–18). The CMR population is relatively stable (14) with an annual change-over rate of 5% of patients. Each patient is assigned a unique identifier number to allow recording and analysis of morbidity on an individual basis (‘medical life histories’). The database also provides individual-specific information on age, gender, socioeconomic status (SES) and the calendar dates on which the patient entered and (if applicable) left the practice. As the GP serves as gate-keeper for secondary care and medical specialists report back to the GP after a referral, the CMR database also includes diagnoses made by any other physician after referral (for this study: respiratory diagnoses made by chest physicians, internists or pediatricians). When an individual leaves the practice, the morbidity registration terminates. The GPs of the four practices meet regularly to discuss classification and coding issues to assure uniformity and accuracy. These factors allow reliable patient tracking and solid opportunities for studying disease longitudinally (16–18).
Morbidity data and diagnostic criteria
Since the start of the CMR, the same standardized classification system has been used to record morbidity. The classification system has been adapted over the years to definitions of the International Classification of Health Problems in Primary Care (ICHPPC)-2 (19) and to other primary care references [for this study: the Dutch College of General Practitioners’ guidelines of asthma (20–22) and AR (23)]. SES is defined according to profession and education level based on the Dutch Standard Classification of Occupations, 1992 (24).
Morbidity definitions and subject selection
The definitions used in the CMR practices to record AR, asthma and eczema are based on the ICHPPC-2 (19) and are shown in Table 1. Using these definitions, two groups of subjects were selected from the CMR database:
|Allergic rhinitis (R97)||Includes hay fever and nasal allergy|
|Asthma (R96)||Recurrent episodes of reversible acute bronchial obstruction with wheeze/dry cough; or diagnostic test meeting currently accepted criteria for asthma|
|Eczema, atopic (S87)||Pruritic exudative lesion with/without lichenification over face and neck, wrists and hands, chest, back of knees and front of elbow|
- 1 Allergic rhinitis group: all patients with a diagnosis of AR in the CMR database in the period between the first of January 1967 until the first of January 2006. Patients older than 50 years at the time of their initial diagnosis of AR were excluded to avoid misdiagnosis of asthma for chronic obstructive pulmonary disease (COPD).
- 2 Control group (‘controls’): subjects without a recorded diagnosis of AR in the database, matched for age, gender, SES, general practice in which they were listed and date of diagnosis.
Subjects with an asthma diagnosis at the start of the follow-up were excluded for the prospective analysis, as these patients already had been diagnosed with asthma and therefore could not develop asthma during the follow-up.
Follow-up of patients in the AR group started at the date AR was diagnosed. For each control subject, a dummy ‘date of diagnosis’ was used that equaled the date of AR diagnosis of the patients with AR to whom they were matched. This fictitious ‘date of diagnosis’ meant that controls had to be present in the CMR database at the same moment in time when an AR diagnosis was given to their matched patient with AR (for instance, a patient diagnosed with AR in July 1987 was matched to a control subject that was also present in the CMR database in July 1987). This accounted for any possible time period effect in diagnoses of AR and asthma. The dummy date of diagnosis marked the start of the observation period for the controls. Follow-up ended (i.e. data were censored) both for patients with AR and controls at the date asthma had been diagnosed, the date the patient left the CMR practice or the selected date of study termination (1 January 2006), whichever occurred first.
The data were analyzed with spss 12.0 for Windows (SPSS Inc., Chicago, IL, USA) and sas 9.1.3 (SAS Institute Inc., Cary, NC, USA). To create a control group that matched the group of patients with AR, propensity score matching (25) was used. A logistic regression model was used first to predict the propensity of having AR using individual characteristics as gender, age, SES, general practice and date of diagnosis. After balancing covariates in the propensity score model, a greedy matching algorithm was used with a 1 : 2 matching from best to next-best for the outcome model. Best matches were defined as those with the highest digit match (0.00001) on the propensity score. The algorithm proceeded sequentially to the lower digit match. The lowest allowable digit match was 0.1. After the first match for each patient with AR had been selected, the same procedure was used to match controls that were not previously matched to get a 1 : 2 matching rate. To examine the extent to which the matching procedures resulted in comparable samples in terms of individual characteristics, chi-square tests were used for categorical variables and Student’s t-tests for continuous variables. P-values used in these analyses were two-tailed and differences with P-values <0.05 were considered statistically significant. A chi-square test was also used to compare the percentage of patients with asthma in the AR group and the control group cross-sectionally.
To take into account the variable follow-up time in the study population, Cox proportional hazard analysis was used to assess the risk of a diagnosis of asthma in patients with AR relative to controls. The Cox proportional hazard model takes maximum advantage of each subject’s available data even when subjects were tracked for different lengths of time, and the outcome of interest might not have occurred. The hazard ratio from the multivariable Cox proportional hazard model was adjusted for the effects of age, gender, eczema, socioeconomic class and general practice. We used ‘physician-diagnosed asthma since start of follow-up’ as the dependent variable. P-values as well as 95% confidence intervals (95% CI) were calculated for the hazard ratios.
We identified 2279 patients with AR in the CMR database. Every patient with AR was matched by the propensity score matching procedure to two controls, which resulted in a control group of 4558 subjects. The total study population consisted therefore of 6837 subjects. Of this group 346 patients had a diagnosis of asthma prior to the start of follow-up (198 patients with AR and 148 controls). As these patients already had asthma and therefore had no chance on developing asthma during the follow-up, they were excluded from the prospective analysis. The exclusion of this group led to a final study population of 6491 subjects (see Fig. 1). The mean age at the start of follow-up was 25.1 years (SD 11.8, range 0–88 years) for the patients with AR and 25.2 years (SD 12.6, range 0–60 years) for the controls (Fig. 2). Table 2 shows the characteristics for the two groups. Compared to the controls, patients with AR had a slightly higher SES and had statistically significant more diagnoses of atopic eczema (19.3%vs 13.0%, P < 0.001). There were no other statistically significant differences. The mean follow-up of the total population was 8.4 years (SD 7.7).
|Total (n = 6491)||Patients with AR (n = 2081)||Controls (n = 4410)||P-value|
|Female||55.2 (3584)||55.3 (1151)||55.1 (2432)||0.9|
|Age in years (SD)*||25.2 (12.3)||25.1 (11.8)||25.2 (12.6)||0.71|
|Low||39.7 (2582)||38.1 (793)||40.5 (1787)||0.04|
|Middle||45.4 (2949)||45.6 (949)||45.3 (1997)|
|High||14.9 (965)||16.3 (339)||14.2 (626)|
|Follow-up period in years (SD)||8.4 (7.6)||8.4 (7.5)||8.5 (7.6)||0.6|
|Eczema†||15.1 (978)||19.3 (402)||13.1 (576)||<0.001|
Association between AR and asthma
Cross-sectional analysis of the data showed that a total of 574 (8.3%) patients were diagnosed with asthma ever (before and after start of follow-up), 15.6% (n = 356) of the AR group and 4.8% (n = 218) in the control group (P < 0.001). Of the 356 patients in the AR group that were diagnosed with asthma, 44.4% (n = 158) were diagnosed after the diagnosis of AR. Of the 218 controls that were diagnosed with asthma, 32.1% (n = 70) were diagnosed after the start of their follow-up.
After start of follow-up, the proportion of newly diagnosed asthma in patients with AR was 7.6% (n = 158) compared to 1.6% (n = 70) in controls (Fig. 1). Chi-square testing showed that this difference was statistically significant (P < 0.001). For patients with a new diagnosis of asthma, the mean duration between start of follow-up and diagnosis of asthma was 5.7 years for patients with AR and 5.8 years for controls (P = 0.9).
Table 3 shows the results of the multivariable Cox proportional hazard analysis. Having been diagnosed with AR was a statistically significant risk factor for asthma (hazard ratio: 4.86, 95% CI: 3.50–6.73, P < 0.001, controls as reference). This difference in risk is depicted in Fig. 3. In time, the cumulative hazard of patients with AR for development of asthma increased statistically significant compared to the controls (P < 0.001). Atopic eczema had no statistically significant effect on the risk for asthma, but age did have a small statistically significant effect on the risk of developing asthma in the future: the younger a subject was at the start of the observation period, the higher the risk of being assigned an asthma diagnosis by their GP (hazard ratio 0.98, 95% CI 0.97–0.99, P = 0.001, 1 year older age as reference) in subsequent years.
|Variables in the Cox proportional hazard model||Reference category||Hazard ratio*||95% CI||P-value|
|AR × eczema||No rhinitis × eczema||0.88||0.46–1.70||0.71|
|Age at start of study||1 year older||0.98||0.97–0.99||0.001|
|SES middle||SES low||1.05||0.80–1.38||0.72|
|SES high||SES low||0.75||0.47–1.19||0.22|
In this primary care study, we found that physician-diagnosed AR is an independent risk factor for a future diagnosis of asthma. Having been diagnosed with AR increased the risk almost fivefold for an asthma diagnosis later in life.
As far as we are aware, this is the first study that has prospectively investigated the association between physician-diagnosed AR and asthma in a primary care population with such a wide age range, such a large study population size and this length of follow-up. As AR and asthma are generally diagnosed and treated in primary care, it is important to assess their relationship in this setting. The Dutch health care system is well suited for primary care research, because all patients are registered with a GP who is the starting point of the health care provided (GP is the ‘gate-keeper’), with specialist care only available after referral. Everyone is insured by a single payer source, and the population is relatively stable. These factors allow reliable patient tracking of diseases over time.
This study used physician-diagnosed AR and asthma that had been recorded at the time of presentation by the patient. The strength of this is – compared to questionnaire-based surveys – that this includes a medical-professional clarification and interpretation of patients’ perceived signs and symptoms. The diagnostic criteria of AR and asthma in the CMR have been defined in line with international criteria (20, 22, 23). The GPs of the CMR network meet regularly to discuss classification and coding issues, and earlier studies have established a high accuracy of diagnosis for a variety of clinical conditions (17, 26). For that reason, we are confident that the quality of the recorded data in this study was at least as good as that of questionnaire-based studies.
Allergic rhinitis can be defined using ‘subjective’ symptoms (blocked or runny nose, sneezing, itchy nose), or using objective findings (e.g. skin prick tests or measurement of IgE-specific antibodies), or a combination of both objective and subjective findings. In this study most AR diagnoses were based on subjective symptoms. Although results of skin prick tests and IgE can be better reproduced, it measures sensitization and not necessarily the presence of relevant illness (27–29). It has been reported that the proportion of IgE-mediated allergy in patients with ‘allergic’ rhinitis as assessed by questionnaire is about 50% (28), while non-AR seems to account for 30–70% of patients with chronic or persistent rhinitis (27, 30). Nevertheless, most studies on AR have been based only on an assessment of symptoms using questionnaires (5, 10, 12, 13, 31). Recall bias and patients’ misinterpretation of questions can influence the results of questionnaire studies (32–34). Probably in epidemiologic studies, there is an overestimation of AR when using questionnaires only.
As AR is a relatively common disorder for which over-the-counter medication is available, many patients will first try to solve their problems themselves before consulting a physician with their symptoms (35). Therefore, it is possible that some of the patients with AR in the general population did not consult their GP and were therefore not included in this study – in particular, patients with mild symptoms. This means that we cannot exclude the possibility that some control subjects did in fact suffer from (mild) AR. Furthermore, the results are not generalizable for patients older than 50 years as we excluded patients above this age at the time of their initial AR diagnosis to avoid misclassification of asthma and COPD in the outcome assessment.
We used the propensity score matching procedure (25) to create a control group and were able to match every patient with AR to two controls. However, as with matching on individual characteristics, this method can only reduce bias arising from the measured characteristics (i.e., age, gender, social economic status, general practice and date of AR diagnosis). The distribution of SES categories differed statistically significantly between the AR and control groups, but was only minor in magnitude. It is, therefore, unlikely that the difference in SES distribution has had a relevant impact on the observed association between AR and asthma, especially while SES was also included as a covariate in the multivariable Cox proportional hazard model.
Several studies have been conducted to elucidate the relation between AR and asthma. Most studies reported findings in line with ours in showing an increased risk of asthma in subjects with AR. However, most of these studies were difficult to translate to the longitudinal observations of patients in general practice, as they were cross-sectional in design, followed up their subjects only for a short period of time or were not based on a primary care population with a wide age range (10, 11, 31, 36). Shaaban et al. (9) carried out a longitudinal population-based study in 14 countries. They reported that AR increased the risk of asthma almost fourfold, but also that non-AR significantly increased this risk. Their population did not include children and given the high prevalence of AR and allergy in childhood, the follow-up from childhood onwards is in our view a particular strength of our study. In their study, Burgess et al. included 7-year-old children and followed them up until the age of 44. The results of this study show that self-reported AR increased the risk of asthma after childhood (37) Settipane et al. (10) carried out a 23-year follow-up study of college students (predominantly males) aged 18–19 and found almost a threefold increase on asthma risk. Huovinen et al. (31) followed up a cohort of twins for 15 years and assessed by questionnaire whether or not the participants had AR. They found an asthma incidence rate ratio of 4.3 for men and 6.0 for women. Linneberg et al. (11) reported that AR predicted development of allergic asthma, but their definition only included asthma symptoms provoked by allergens. Porsbjerg et al. (12) found that presence of AR as assessed by questionnaire did not increase the risk of asthma in 291 atopic and nonatopic subjects who were followed-up for 12 years. Univariate analysis showed a nonstatistically significant odds ratio (OR) of 1.7 (95% CI 0.7–3.9). These findings are partly in line with reported findings by Plaschke et al. (13), who found that atopic subjects (assessed by skin prick testing) did have an increased risk of asthma, whereas nonatopic subjects did not (OR = 5.7, 95% CI 2.2–14.6).
Next to the epidemiologic evidence of the relation of AR and asthma, there are also pathophysiologic data to link rhinitis and asthma. Anatomically, the nose and lungs are closely related and both rhinitis and asthma are manifestations of an inflammatory process within this continuous airway system (2, 38, 39). At this moment, it is unknown whether AR is an early clinical manifestation of allergic disease in atopic subjects which will later progress into asthma or whether AR itself is an actual risk factor for asthma. In this study, we found that 44% of patients with asthma were diagnosed with asthma after the diagnosis of AR and 56% before or at the time of the diagnosis of AR. This suggests that there is a link between AR and asthma but not necessarily a prospective one in every case.
Clinical implications and conclusion
Our study shows that patients with physician-diagnosed AR had a fivefold increased risk of subsequently developing asthma compared to controls without AR. The findings imply that patients who are diagnosed with AR deserve extra attention from their GP with regard to possible coexistence of asthma or its development in subsequent years. We conclude that AR appears to be a risk factor for a diagnosis of asthma in this primary care population.
Department of Primary and Community Care, Radboud University Nijmegen Medical Centre.
Conflict of interest
The authors declare no conflict of interest.