Discussion
The main finding of these prospective population-based cohort studies was that the incidence of asthma among adults in Northern Sweden was stable from 1996 to 2016, and at similar level as from 1986 to 1996 [12,15]. Thus, the reported increase in asthma prevalence during the last decades is not explained by an increase in asthma incidence, but by the stable and relatively high incidence, 2-3/1000/year. The incidence was consistently higher among women than men. In analyzes adjusted for covariates, allergic rhino-conjunctivitis was the main risk factor for incident asthma both 1996 to 2006 and 2006 to 2016.
To the best of our knowledge we are the first to report about trends in asthma incidence among adults over several decades. A review article including incidence studies from 1950s to 1990s indicated an increase in incidence rate by time, however, the methods varied substantially between the included studies [28]. Three longitudinal population-based studies performed in Sweden and neighboring countries during the 1980´s and 1990´s with almost identical methods found incidence rates of 2-3/1000/year [15,16,18]. Our results, 2.4/1000/year from 1996 to 2006 and 2.6/1000/year from 2006 to 2016, are in line with these studies, including the previous study in our region (Figure 1) [15], and indicate a continuing stable trend in incidence rate in Sweden as a whole. However, an incidence rate of 2-3/1000/year in combination with low remission of asthma in adulthood [6,7,18] and similar mortality among adults with asthma as in the general population, means about 1% unit increase in prevalence every ten years and thus contribute to the slowly increasing prevalence of asthma among adults.
Only a limited number of studies about the incidence rate of asthma in adulthood are based on prospective longitudinal population-based studies [12-21,23], while others are based on registers or retrospective estimates [22,24]. These studies show considerable variation in results, 0.7-6/1000/year due to variations in the study design including the age of the population [11,13,14,25], and the definitions of the outcome and the population at risk [12, 14-16,18]. Studies ending up with high incidence rates have often not excluded individuals with symptoms common in asthma from the population at risk. To enable valid comparisons between different studies it is paramount to use identical methods.
Symptoms of asthma may occur several years before a diagnosis of asthma is made and these symptoms may reflect undiagnosed asthma in the population, why the incidence rate measured prospectively may be overestimated. To reduce this bias it is reasonable to also exclude subjects reporting symptoms of asthma from the population at risk. When comparing different populations at risk (model A vs B) we found that the difference in incidence rate of physician-diagnosed asthma between the two models was much smaller in the two current cohorts under study compared with the study from the 1980s (Figure 1). When using model B (not correcting for possible under-diagnosis), the incidence of physician-diagnosed asthma was 4.4-4-8/1000/year in the current cohorts, while it was 8/1000/year in the 1980s cohort [12]. This may reflect a decrease in under-diagnosis of asthma since the 1980s, probably due to changed diagnostic practice and higher awareness of asthma in healthcare and society today [2,26]. However, at the follow-up of our two current cohorts in 2006 and 2016, individuals with incident asthma reported similar prevalence of symptoms of asthma and use of asthma medication. This indicate a similar disease burden among individuals with adult onset asthma today as ten years ago, which in part contradicts the suggested diagnostic drift regarding asthma in healthcare. On the other hand, the major change may have occurred before the millennium shift as a consequence of a Swedish national program focusing on asthma during 1995, and perhaps even related to the start of the OLIN-research program in the study area in 1986, resulting in increased awareness of asthma in the society. Still, the higher incidence rate of use of asthma medication compared with asthma diagnosis may indicate some under-diagnosis of asthma also today. However, asthma medication is also used in other respiratory diseases, such as bronchitis and COPD, which may contribute to the high incidence.
In line with previous studies, we found that women had a higher incidence rate of asthma compared with men [12-24,29,30] (Figure 2). This is in contrast to children, where the incidence is higher among boys than girls [8,9,11,22]. The reason for the sex differences is unclear but may be related to hormonal status [29]. The obese-asthma phenotype, often found in adult onset asthma among women [5,29], may be mediated by interactions between sex hormones and systemic inflammation [31]. However, female sex has been reported to be a predictor for persistent asthma from childhood to adulthood [32,33] and it seems that the increasing prevalence of asthma among adults is mainly driven by women, i.e. by lower remission rate of childhood asthma and higher incidence rates of asthma in adult age in women than in men.
Allergic rhino-conjunctivitis is a well-known risk factor for asthma [4,17,19,21] and is a reasonable proxy for allergic sensitization, at least among younger and middle-aged [21]. In the current study it was the strongest and most stable risk factor for asthma which further strengthen the causality of the association. Several previous studies have found smoking or ex-smoking associated with incidence of asthma [4,10,12,15,19,23]. In the current study we found that, from 2006, neither smoking nor ex-smoking remained as risk factors for incident asthma among adults. This change is probably a result of the considerable decrease in smoking that have occurred during the last two decades in Sweden [2,26]. Not only the prevalence of smoking has decreased but, also the number of cigarettes smoked per day has been reduced [2]. Thus, the total level of tobacco exposure had decreased significantly which is beneficial also for the general public health. In contrast to previous studies which have reported several markers for low socioeconomic status to be associated with incident asthma [34-36], we did not find any consistent associations with socioeconomic status based on occupation. However, though not significant, the risk ratio for manual work was 1.75 in the 1996 cohort while the RR was close to 1 in the most recent cohort. This may be due to improvement of the working environment, but also due to lack of power in our study, not allowing more detailed occupational classifications.
Several strengths support the validity of the results. First, we used two large population-based cohorts within the same age-span and geographical area and both cohorts were followed for ten years. Second, the response rate was high in both cohorts, both at recruitment and follow-up. Third, for high comparability with previous studies [12,15,18] we used identical methods in the calculations of incidence, and these were based on the same validated questionnaire [27]. Furthermore, the prospective design is, compared with retrospective studies, less associated with recall bias [11,25]. Although the internal validity can be regarded as high, the external validity is unclear due to the lack of studies of trends in asthma incidence. A limitation worth noting is that despite the large sample sizes of the cohorts and the relatively long follow-up, the study lacks power for subgroups analyses. The studies were based on questionnaire data which may have introduced some bias, why we cannot exclude that some of those diagnosed as having asthma in reality had COPD. However, the incidence were similar across all age groups. Another weakness is the limited information about risk factors, such as body mass index and education level, factors that have been associated with adult-onset asthma [4,5,29,34].