Associations between socioeconomic position and asthma: findings from a historical cohort

The Glasgow Alumni Cohort has been described in detail elsewhere [15]. Briefly students attending Glasgow University from 1948 to 1968 were invited to attend a health examination at the Student Health Service. Medical history, sociodemographic and socioeconomic characteristics and physical examination were recorded through a physician-administered standardised questionnaire. Medical history included conditions such as bronchitis, asthma, hay fever, eczema/urticaria and questions on health behaviours. Data was collected on 15,322 students of which 11,756 were men, representing approximately 50 % of the student population. Students who were more than 30 years old at the time of examination were excluded (n = 482) leaving 11,274 participants.

Since 1998, 82.5 % of the total Glasgow Alumni Cohort has been successfully traced through the National Health Service Central Register. Between 2001 and 2002, members of the cohort that had been traced were asked to complete a postal questionnaire to collect additional early life and adulthood information and respiratory and other chronic diseases. Completed questionnaires were returned by 4,101 participants. Ethics committee approval was received for the study.

The early life outcome variables included were previous medical history of asthma, eczema/urticaria, hay fever and bronchitis measured at University entry through a physician-administered standardised questionnaire. These were categorized as asthma alone (asthma without eczema/urticaria or hay fever), asthma with atopy (asthma with eczema/urticaria and/or hay fever), and atopy alone (eczema/urticaria and/or hay fever without asthma). Eczema and urticaria were asked jointly in the questionnaire and could not be separated in the analysis. To evaluate whether this had an effect on the socioeconomic patterning of atopy, additional analysis were carried out defining atopy as reporting of hay fever only. Asthma ascertained in the adult postal follow-up was categorized as asthma ever (doctor diagnosis of asthma ever) and adult-onset asthma (doctor diagnosis of asthma ever combined with age of diagnosis greater than 30 years of age). Bronchitis was ascertained in the adult postal questionnaire if the participant responded positively to the question ‘Have you ever been told by a doctor that you have, or have had, bronchitis?’

Early life SEP was based on father’s occupational class (collected at university entry questionnaire) and classified on a five point scale from I (professional occupation), II (managerial and technical job), III (skilled non-manual and manual job), VI (partly skilled manual job) and V (unskilled manual jobs) using the Registrar General’s Classification, the official UK social class classification at that time (and until 2002) [16]. Due to small numbers social class groups III, IV and V were combined into one single group. The postal adult follow-up questionnaire included recall of early life SEP based on father’s main occupation, which was classified following the same standardized criteria into I–V groups, number of household amenities when starting university (included fridge, telephone, car and television), adult SEP based on participants main occupation (I–V), adult household crowding, income, home ownership and car ownership.

Other variables collected at University entry included birth order, number of siblings, smoking status and family history in first degree relatives of asthma, hay fever and eczema. Body Mass Index (BMI) was calculated from height and weight measured at University entry and used divided into tertiles to form a categorical variable as there were very few participants in the BMI > 30 when using WHO classification.

Statistical analysis was completed in Stata 12.1 (Texas, USA). Chi-squared test was used to describe the association between (1) early life SEP and respiratory and atopic diseases and, (2) adult and early life SEP (as reported from the postal follow-up questionnaire) with adult respiratory diseases. We present the p value for trend using the exposure as a continuous variable. Multinomial logistic regression modelling was performed to determine the association between early life SEP and a 4-group outcome variable including asthma only, atopy only, asthma with atopy, and no asthma or atopy as the reference category. Results from these models are presented as multinomial odds ratios (MOR) [17, 18] although they are often referred to as relative risk ratios in the literature [19]. Multinomial logistic regression models were first adjusted for age and year of survey to investigate possible confounding by age of participant or change of definitions over time of surveys. Likelihood ratio tests (LRT) were used to test whether additional variables contributed to the model. The covariates considered for the adjusted model were; BMI, birth order, number of siblings and smoking status. The MOR and 95 % confidence intervals for the linear trend was determined by entering early life SEP as a continuous variable into the multinomial logistic regression model. Logistic regression models were used to investigate whether the association between asthma and early life SEP differed according to atopy status. The interaction was formally evaluated through a LRT comparing models with and without the interaction term. Logistic regression models were used to investigate the association between early life SEP and early life asthma classifying the latter as asthma yes/no (ignoring atopy status) and equal to adult asthma.

Finally, logistic regression modelling was performed to determine the association between early life SEP and adult onset asthma and asthma ever using data from the adult follow-up.

In this historical cohort of male university students, we found that a different socioeconomic patterning was already present before the rises in the prevalence of asthma and atopy occurred in the UK. Asthma alone was higher in students from lower socioeconomic background whilst lower SEP was associated with lower odds of asthma with atopy and atopy alone. In addition, number of household amenities when starting university was associated with higher risk of adult onset asthma and bronchitis ever however this was attenuated when adjusting for age. Adult SEP was not associated with adult onset asthma.

The prevalence of asthma with atopy and asthma alone in this cohort was 1.25 and 2.38 % respectively and is comparable to asthma prevalence reported in the University of Wales between 1950 and 1954 [20]. Graham et al. [21] reported a higher prevalence of childhood asthma in higher social classes and a higher frequency of history of asthma in non-manual classes compared to manual classes were reported in a national study of childhood asthma [22]. However wheezing is reported to be lower in higher social classes in the Schoolchild Chest Health Survey (1966) [23] and no socioeconomic patterning was found in the 1958 birth cohort [24]. Asthma severity in childhood has been more consistently associated with lower childhood SEP [25, 26]. Conflicting results in the direction of association between early life SEP and asthma may be due to the inability to separate asthma with or without atopy phenotypes in previous studies. Pearce et al. [27] have suggested the importance of considering asthma with atopy and asthma alone separately in a bid to assess different aetiological factors (other than atopy) in the development of asthma. In this cross-sectional study although asthma and atopy may co-exist we are not able to determine if atopy is causally related to asthma. However, the different socioeconomic patterning of these two phenotypes of wheeze and asthma also points to different aetiological mechanisms each relating to environmental exposures and/or disease awareness that are differently distributed by socioeconomic groups.

There is only one study that has previously reported an association between early life SEP and asthma according to atopic status and from a period preceding the well-documented rise in asthma prevalence in Sweden [14]. Bråbäck et al. [14] investigated the associations of SEP and asthma in cohorts born between 1952 and 1977 and reported low SEP was associated with a higher risk of asthma without rhinitis across all cohorts. On the other hand, asthma with rhinitis, initially more common in higher social classes became more prevalent in the lower groups due to a faster rise among the latter. Non-atopic wheeze (as categorised by IgE status), but not atopic wheeze, has also been shown to have a higher prevalence in lower social classes (as categorised by head of household occupation) in the Health Survey for England (ages 11 and over) [28].

Similarly, there are a limited number of studies that have looked at the associations between early life SEP and hay fever and eczema within the similar time period as our cohort. Eczema, diagnosed by a school medical officer, and self-reported hay fever was higher in higher social classes in a cohort of children born in 1958 [24]. Parentally-reported eczema in childhood (age 5 or earlier) has also been associated with higher parental education in a later cohort born in 1970 [29].

On the other hand, a higher prevalence of asthma with atopy and atopic conditions may be due to over-reporting among students from higher SEP (or under-reporting among low SEP students) who may have had greater awareness of these conditions and greater access to medical services. Although this has been previously suggested [21, 30] whether this is the case has not been established and a report from the 1958 birth cohort suggested that the over-reporting of eczema was among low and not the higher SEP groups, therefore implying an even greater differential between socioeconomic groups [24].

Results from Bråbäck et al. [14] showed the steepest increase of asthma with rhinitis was seen in the lowest SEP group indicating that perhaps changes in exposures or awareness were greater in this group. If the role of SEP has changed over time, and cannot be explained by changes in disease awareness, suggests that exposures that lead to allergy and sensitisation may have become similar across all SEP and explain the inconsistency in associations in contemporary studies.

Our findings for the association between birth order with asthma and hay fever have been reported previously in this [31] and other studies [5, 32]. Both asthma with atopy and atopy alone, were more prevalent in lower birth orders (first or second born). Although a trend of lower prevalence with lower birth orders was also present for asthma alone the magnitude of differences was smaller and it could be due to chance. Thus, it is not clear whether the two asthma phenotypes have different associations with birth order in this cohort. We could not test a previously reported interaction between birth order and SEP [31] due to the small numbers and wide confidence intervals when classifying asthma according to the presence of other atopic conditions. Higher number of siblings, in the other hand, was only associated with lower prevalence of hay fever and bronchitis.

In our analysis we have used a 4-category outcome that included asthma only (asthma without eczema/urticaria or hay fever), asthma with atopy (asthma with eczema/urticaria and/or hay fever), and atopy only (eczema/urticaria and/or hay fever without asthma) compared to no asthma and no atopy. Multinomial models allow analyzing the three outcome groups simultaneously against the same reference level and allow testing for statistical significance accounting for the probabilities of all response categories. The group characterized by “asthma with atopy” is however difficult to interpret as it includes subjects for whom the asthma is related to their atopic status and asthma that simply coincides with atopy [33]. Many studies in this area have used different reference groups often making comparisons across studies difficult and the debate of which categories of asthma and atopy to use is on-going [33, 34]. Stratifying for atopy status allows to formally test whether the association between asthma and early life SEP differs across atopy groups [33].

Follow-up data from this cohort demonstrated no association with early adulthood or adult SEP with doctor’s diagnosis of ever asthma. However doctor’s diagnosis of adult-onset asthma and bronchitis were more prevalent among those living in households with fewer amenities during early adulthood. This was attenuated when adjusting for age. We could not differentiate asthma types in adulthood according to atopic status as there was no information collected on eczema/urticaria and hay fever in the adult follow-up postal questionnaire. Other investigators have reported a higher prevalence of adult asthma in lower adult socioeconomic groups [11, 35] but only a very small number of students were of mid-low adult social class in this cohort (90–95 % of students were of class I and II in adulthood).

A major strength of this study is the unique historical timing of the cohort, before the increase in asthma and atopic conditions, and before widespread use of antibiotics and immunisations. We could separate types of asthma according to the presence (or absence) of other atopy-associated conditions and measured early life SEP at the same time. Although medical history data was collected by physicians we cannot rule out misclassification of disease. The strong correlations between asthma and bronchitis highlight the difficulty to differentiate these diagnosis.

Diagnosis of asthma and bronchitis are likely to have overlapped. Indeed, bronchitis appeared to have mixed characteristics with asthma with or without atopy, namely higher prevalence in the leanest groups, lower birth orders, lower number of siblings and family history of asthma and family history of eczema and/or hay fever (family history of bronchitis was not available), suggesting that a degree of misclassification occurred between these two conditions. Atopy was also more frequently reported among students with bronchitis. On the other hand, in a previous report from this study we found a higher mortality in adulthood due to respiratory conditions among the students reporting asthma or bronchitis in early adulthood but only history of bronchitis was associated with higher mortality due to cardiovascular disease (as previously reported in the literature) suggesting some degree of ability to differentiate between these two conditions [36].

Members of this cohort were students at the University of Glasgow from 1948 to 1968 and therefore are a select group of individuals who could afford high education [15] with the majority of participants indicating father’s social class to be I, II and III. Most participants to the survey had professional occupations in their adult lives (90–95 % in class I and II). This is likely to have limited the socioeconomic variability in adult asthma and may limit the generalisability of these results. However, the associations reported with early life SEP and types of asthma are less likely to be confounded by adult socioeconomic circumstances. Thus, this relative homogeneity of adult socioeconomic circumstances becomes a strength when evaluating the association with early life SEP.

Only men are included in this report because of the small number of women attending university from 1948 to 1968. There were no associations between early adulthood SEP and asthma with atopy, asthma without atopy, hay fever only, eczema/urticaria only, in women in this cohort (results available from the authors) but we cannot rule out lack of power to detect this association in women. Ethnicity was not recorded because most of students attending university at that time would have been white. Similarly, potential relevant explanatory variables, such as parental smoking or access to health care, were not recorded and could not be evaluated.

Although a great proportion (82.5 %) of the Glasgow Alumni Cohort was traced through the National Health Service Central a smaller proportion responded to the follow-up questionnaire. Those who did not respond were more likely to be smokers though there were no other differences between responders and non-responders. Although this could present a potential bias to our results this would only be the case if the results in the non-response group demonstrated an opposite social patterning of asthma. Moreover we have previously reported that the association of early-life weight (reported in the Student questionnaire) and type-2 diabetes onset in adulthood (from follow-up questionnaire) was similar to another cohort with a greater response rate (90 %) [37].