EPIDEMIOLOGY OF ASTHMA AND ALLERGY

Contents

This document was last modified on 10-March-2009.

Descriptive epidemiology of asthma

Incidence and prevalence in Australia

A number of studies offer information on current incidence and secular trends for asthma. The 1989-90 National Health Survey [ABS 1991] gave information on self reported asthma in a probability sample representing the entire population. Recent asthma (in the previous two weeks) was reported by 4.3%; another 4.2% reported that they had previously suffered asthma as a long term condition (6 months or more). Broadening the definition to include a "wheezy or whistly chest" gave a lifetime prevalence of 23.8%.

For reported asthma, the sexes were almost equally affected (51% male). There had been a marked increase in reported long-term asthma at all ages since the 1977-78 Survey, most noticeably in those under 15 years of age, where the prevalence had risen from 4% to 14%. An examination of birthplaces noted the (age-sex adjusted) prevalence of asthma was highest in those born in Australia and New Zealand, next for those from the UK and Ireland, less for those born in Southeast Asia, Africa and Southern Europe, and least for those born in Western Europe.

Table 1. Cumulative incidence of asthma (either long term or recent) in the 1989-90 Australian National Health Survey.
Age0-1415-2425-3435-4445-5455-6465-7475+
Cum Inc (%)13.910.46.95.75.85.76.84.2

The 1995 Australian National Health Survey found that the self-reported prevalence (recent plus long term) rose further from 8.5% to 11.3% of the population [AIHW 2000], and in those 5-14 years of age rose to 19.2%. The Australian Burden of Disease and Injury Study [1999] placed asthma as the fourth commonest condition by prevalence and twentieth by incidence, making it the ninth leading cause of disease burden (2.6% of DALYs; 4.5% of hospital separations; A$478 million annually in 1993).

In 1981, Woolcock et al [1987] performed bronchial challenge on a random subsample of all adults taking part in the Busselton (WA) health study. Out of the estimated 6000 eligible adults in the town, 3590 took part in the 1981 round. One in four (N=922) were randomised to complete a respiratory questionnaire (based on the MRC respiratory questionnaire, supplemented by specific questions on wheezing). Subsequently they underwent histamine challenge test (HCT), and skin prick testing for 14 common allergens. Only five eligible subjects refused to take part. In 1990, a further sample of 5020 out of 9000 subjects returned a screening questionnaire, of whom a stratified random sample of 1028 completed another respiratory symptoms questionnaire (this time based on the IUATLD questionnaire) and underwent HCT [Peat et al 1992]. This age of members of this group was limited to 18- 55 years. It should be noted that no information is available on how many participants took part on both occasions.

Restricting results to those aged 18-55 years, a doctor diagnosis of asthma was reported by 9.0% in 1981, and 16.3% in 1990. The lifetime risk of wheeze in this population was 17.5% in the first sample, rising to 28.8% of the second sample. However, there was no associated increase in either the prevalence of skin atopy or bronchial hyperresponsiveness (see below). This meant that the narrowly defined diagnosis of current symptoms plus hyperresponsiveness did not alter between the two waves.

Robertson et al [1991] have reported a survey of childhood asthma in Melbourne. Schools were randomly selected, and all children aged 7, 12 and 15 years in that school surveyed. Responses to a respiratory symptoms questionnaire were received from 9192 of 10198 (90%). Wheeze in the previous twelve months was reported by 23% of 7 year olds, 22% of 12 year olds and 19% of 15 year olds (though amusingly the rates of exercise induced wheeze were 15%, 20% and 23% respectively). The prevalence of doctor diagnosed asthma was fairly constant across the age groups (24%, 23%, and 22%). Our findings in a survey of 2279 Australian twin pairs aged 4-13 years are very similar - wheeze in the previous twelve months, 24%; doctor diagnosis, 26% [Duffy et al unpublished], with little variation across the age range.

Comparison with other countries

Several hundred prevalence studies of asthma have been performed this century [see Gregg 1983]. I will only review some recent results and make some general comments. First, asthma is more a disease of the developed world than the Third world. Secondly, the incidence (as well as mortality) of asthma seems to have increased this century in the developed world, and especially in the last 20-30 years [Buist 1989; Robertson et al 1991; Burr et al 1989].

Gregg [1986] has divided countries by their reported asthma rates into high, medium and low incidence countries. High incidence countries include Australia and New Zealand. Medium incidence countries are USA, Canada, UK and mainland European countries. Low incidence countries include most of Africa and Asia. Scandinavia has been previously regarded as a low incidence region, but along with most of the developed world, has seen a recent increase in asthma.

The United States National Health Interview Survey for 1989 gave the point prevalence of self-reported asthma at 4.8% of whites under age 45 years rising to 5% of those over 65 years [National Center for Health Statistics 1990]. The Health Interview Evaluation Survey [NCHS 1994] suggested there was 20% underreporting of this condition, when validated against health records. Interestingly, the comparable black figures were 5.7%, and 6.9%. This racial difference is only partly explainable by economic, lifestyle and geographic covariates [Turkeltaub & Gergen 1991; McWhorter et al 1991]. The NHANES II (1976-80) estimates of asthma prevalence [Turkeltaub & Gergen 1991] were 6.9% for whites and 9.2% for blacks. Estimates for Canada are similar [Svenson et al 1993].

New Zealand is known to have one of the highest rates of morbidity and mortality due to asthma in the world. In rural adolescents, for example, Shaw et al [1991] reported the (lifetime) prevalence of asthma as 13.7%, with active asthma present in 8.7%. Asher et al [1988], found a prevalence of 14.2% in urban 8-10 year olds, and Sears [1986], 16.8% of urban 15 year olds. In common with the US, racial differences in prevalence exist - Maoris having a significantly higher incidence of asthma. The Tokelaun migration study found that Polynesian children migrating to New Zealand had increased levels of asthma compared to their relatives remaining in a more traditional environment [Waite et al 1980]. Specifically, 7% of children in New Zealand were asthmatic but only 1% of those in Tokelau (sex-adjusted M-H OR=5.8, 95%CI=2.9-11.7). The prevalence of asthma was the same in both Tokelauan children born in New Zealand after the migration of their parents, and in those born on Tokelau itself. Shaw et al [1991] thought the white-Maori difference in adolescent asthma to be explained entirely by differences in smoking rates.

Temporal changes

It is clear that the prevalence of asthma and allergic disease has been increasing in developed countries for the last thirty to fifty years, though this has been most apparent since the 1960's. Asthma mortality data offer the longest continuous series, but interpretation is made difficult by the changes in diagnostic classifications (ICD) over time. For Australia [Taylor et al 1997], in persons aged 5-34 years (thus limiting misclassification), mortality rose from approximately 0.25-0.5 per 100000 person- years prior to 1950 in a linear fashion up to 1.2 per 100000 in 1955, fell briefly, rose to 2 per 100000 in the late 1960's , fell through the 1970's to 1 per 100000, reached a second peak in the 1985 of 1.5 per 100000 and has fallen below 1 per 100000 since then. In age-period-cohort Poisson regressions, there were linear increases in mortality with each 5 year birth cohort since approximately 1930. The only period showing an increase was in the 1960s, with a steady decrease in adjusted mortality into the 1990s, interpreted as due to increased use of inhaled corticosteroids.

For New Zealand, (all-age) asthma mortality rose from approximately 0.5 to 1 per 100000 person-years prior to 1940 in a linear fashion up to 1.5 per 100000 in 1955, fell briefly, rose to 3 per 100000 in the late 1960's (in common with Australia, Great Britain, and the United States), fell through the 1970's, but reached a peak in the 1980's of 3-4 per 100000 [Beasley et al 1990]. Deaths from asthma increased from 1.3 to 1.9 per 100000 in the US in the 1980's alone [MMWR 1992].

Self-reported long term-asthma increased from the Australian National Health Survey 1977-8 figure of 2.4% up to 8.0% in 1989-90 [ABS 1991]. More strikingly, in Melbourne schoolchildren, Robertson et al [1991] found "ever" wheeze had increased in seven year olds from 19.1% in 1964 [Williams & McNichol 1968] to 46% in 1990. In the Busselton study [Peat et al 1992], reported symptoms increased over a nine year period, but there was no matching increase in bronchial hyperresponsiveness or atopy, suggesting this might be an artefactual increase. Cullen et al [1968] however, reported the prevalence in 1966 of doctor-diagnosed asthma (MRC questionnaire) in Busselton was only 3%. In a recent study of childhood asthma in Wagga-Wagga, Downs et al [2000] showed that the increase in diagnosis of asthma is associated with an objectively measurable increase in allergic sensitisation, positive skin prick tests to D pter (house dust mite) rising from 18.6% in 1992 to 30.3% in 1997.

In the UK National Study of Health and Growth [Burney et al 1990], a history of persistent wheeze was present in approximately 2% of children born in the mid 1960's, rising to 4% in those born in the mid 1970's. The prevalence of diagnosed asthma rose 7% with each subsequent (annual) birth cohort for males, and 13% for females. Impressive evidence was presented by Burr et al [1989], who found that the prevalence of diagnosed asthma rose from 6% to 12% from 1973 to 1988 in South Wales. Bronchial hyperresponsiveness to exercise (free running) also rose from 2% (25% fall in PEF) to 4%.

Examinations carried out on large samples of young men called up for military service have also documented changes in asthma using standardised examinations. Aberg [1989] reported an increase in diagnosed asthma in Sweden from 1.9% in 1971 to 2.8% in 1981. In Finnish conscripts, asthma prevalence was under 1% before 1961 [Haatela et al 1990]. Subsequently, it has risen to 1.9% in 1989. In Israeli seventeen year olds, "ever" asthma rose from 7.9% in 1986 to 9.6% by 1990 [Auerbach et al 1992].

Natural history/course of asthma

The incidence of new asthma is highest in children. Allergy can be demonstrated in 90% of childhood asthmatics, and age at onset is negatively correlated with the number of parents with a history of allergic disease [Cooke & Vander Veer, 1924]. The long term course of childhood onset asthma has been outlined by several cohort studies.

Williams and McNicholl assembled a cohort of 378 Melbourne 7 year olds with a history of wheeze and 106 without such a history in 1964 from a larger cross-sectional study. This group has been followed to age 28 years [Martin et al 1980; Kelly et al, 1987, 1988, 1990]. Of the controls, 32% of those followed up at age 28 years had developed wheezing, and 10% had wheezed in the previous three months. Of those who had experienced recent asthma at age 14 (in the previous twelve months, with a lifetime history of more than 5 episodes of wheeze), 68% had wheezed within three months of testing at age 28 years. There was a linear relationship between frequency of symptoms and BHR, eosinophilia, and allergen skin wheal diameter. The sex ratio for recent asthma was 72% (M:F) at age 14, 60% at age 21 years, and 61% at age 28 years. While eosinophil count fell with age, allergen wheals tended to increase in size.

Similar results were reported by Gerritsen et al [1990a, 1990b] and Roorda et al [1992, 1993] who followed 525 (78% of those seen 1966-9, 86% of those seen 1972-6) of the asthmatic children seen at their outpatient clinic 1966-1976. Boys made up two-thirds of the sample, over 95% were allergic to house dust, and half the sample had suffered atopic dermatitis. Sixteen years later, 449 of the sample were successfully reexamined. At mean age 26 years, 43% of the 1966-9 cohort still reported current asthma symptoms, and 69% of the later cohort - though different questionnaires were used. The mean age at remission of asthma in the remainder was 15 years. Only one- quarter reported current atopic dermatitis, but the prevalence of hayfever had risen from 10% to 20%. All but one of the adults was still allergic to house dust, and adults responded to lower doses of house dust extract than they had as children, although the problem of allergen standardisation makes this difficult to interpret. Adults responded to a greater number of allergens than they had as children.

A subset of 31 adults from the 1966-9 group underwent house dust inhalation challenge. A positive response had been recorded for 25 of these subjects as children, and 22 still responded (9 with a late response) - suggestively, 11 of the 13 with current symptoms, and 11 of the 18 in remission (OR=3.5, exact 95%CI=0.6-27.1). Using summed symptom scores, the responsive group had experienced significantly more severe asthma as a child and as an adult.

Descriptive epidemiology of bronchial responsiveness

Although bronchial challenge testing has been in clinical use since the 1940's, it is only in the last 15 years that population surveys of bronchial responsiveness have been carried out. The technique has become more acceptable for epidemiological purposes with the introduction of shorter protocols, as the original clinical protocols can take up to 45 minutes. The method of Yan et al [1983] is the most widely used of the brief protocols (and will be described in the methodological section). Woolcock and associates have used this method to survey mainly children, but also adults, from the representative community samples.

In Australia

In the previously mentioned Busselton survey of Australian adults [Peat et al 1992], significant bronchial hyperresponsiveness was present in 11.4% of the 1981 sample (either PD20ó3.9 umol HIS or positive bronchodilator test - 15% increase in FEV1). Only 68% of this group reported symptoms consistent with asthma, though this proportion increased in severe BHR. A similar pattern was seen for skin atopy, with 32% being unresponsive to any allergens tested. Clinical current asthma (with BHR) was found in 5.6% of subjects overall, and 7.4% of those aged 18-55 years. In the 1990 group, BHR was present in 7.9% of the subjects. The prevalence of current asthma was 7.2%.

Woolcock's group has performed a number of surveys in Australian children in different localities representing the climates of inland, coastal urban and rural Australia [summarised in Peat et al 1992]. These suggest that BHR tends to decrease with age in children (see Table).

Table 2. Prevalence of BHR in Australian schoolchildren in studies using the Yan HCT protocol.
CentreNumber of childrenAge RangePoint Prevalence (%) of BHR(95% CI)
Belmont NSW (1982,4)9937-1015.8(13.5-18.1)
Wagga NSW (1982-3)13717-1019.6(17.5-21.7)
Busselton WA (1983)43811-148.4(5.8-11.0)
Busselton WA (1983)4477-1015.0(11.7-17.3)
Villawood NSW (1986)12177-1215.3(13.3-17.3)
Burra SA (1988)618-1221.3(11.9-33.7)
Kingston SA (1988)1218-1222.0(14.6-31.0)

Comparison with other countries

A number of surveys are available for New Zealand. Pattemore et al [1990] reported a study of 2053 randomly selected Auckland primary school children tested by the Yan protocol. BHR was present in 15.9% (here defined as PD20<7.8 æmol HIS). Only 46% of this group reported ever being diagnosed as asthmatic and/or asthma symptoms in the previous twelve months. Those without symptoms tended to have milder BHR, but significantly, did not have an excess of hayfever, eczema or other atopic conditions. Of those children with a diagnosis of asthma, only 52% had demonstrable BHR. Sears et al [1991] reported on the birth cohort of 784 (of 1037) 11 year old children in the Dunedin Multidisciplinary Health and Development study. Bronchial hyperresponsiveness to methacholine (PC20<25 mg/ml) was found in 15% of the total sample, and in 70% of those reporting current asthma.

Cockcroft et al [1984] examined a random sample of 400 (out of 490 eligible, 82%) Canadian university students. This group, with a mean age of 21, had a prevalence of BHR of 10%.

Burney et al [1987] studied a random sample of adults from the South of England who had taken part in a questionnaire survey of respiratory symptoms. There was a 60% cooperation rate (522/873). Subjects underwent skin prick testing (mixed grasses, D pter and cat) and a histamine challenge using the Yan protocol. Hyperresponsiveness was present in 14% of the subjects - 14% of those aged 18-24 years, 10% aged 35-44, and 24% of those aged over 44 years. This pattern is consistent with the findings in Australian children of an initial decrease with age, and those of several other studies of an increasing prevalence in later life [Sparrow 1989; Burney 1987]. Additional factors increasing the probability of BHR were smoking and skin atopy. In a separate study [Chinn et al 1987] of 106 subjects, the test-retest correlation for log PD20 (FEV1) over 1-14 days was 0.81.

The Vlagtwedde-Vlaardingen study in the Netherlands [Rijcken et al 1993a] is a large cohort study. From 4700 subjects, repeated 25% samples underwent histamine inhalation challenge, so that responsiveness was assessed in 2216 subjects (mean age 40 years). Nonspecific hyperresponsiveness was found in 33% of this sample, and increased with age (OR=1.1 per year of age). In the individuals who underwent repeated challenge (over periods of up to 18 years), 11% were consistently hyperresponsive, and 41% consistently nonresponsive - that is half the group were intermittently responsive [Rijcken et al 1993b]. A change in PC10 of two doubling doses was seen in 21% of subjects over a three year interval, and in 43% over 18 years (a test-retest correlation of 0.4).

Enarson et al [1987] studied methacholine responsiveness in 1392 male workers in three industries. Hyperresponsiveness (PC20ó8 mg/ml) was present in 16.4% of the sample, and increased with age. In another publication [Kennedy et al 1990], they report BHR present in 14% of a subsample of 654 men that excluded all those with a history of asthma.

Sex differences have been reported in some studies. Leynaert et al [1997] performed a methacholine challenge on 799 subjects aged 20-44 years in two centres as part of the European Community Respiratory Health Survey. BHR (defined as a PD20<4 mg/ml) was present in 34% of woman and 12% of men from Paris, and 43% of women and 29% of men from Montpellier. These sex differences became nonsignificant after adjusting for basal FEV1 level, but not for smoking, occupation etc.

Associates of bronchial hyperresponsiveness

Nonspecific hyper-responsiveness is associated with atopy. Cockcroft et al [1984] found a correlation of -0.34 between a four point atopy score, and bronchial responsiveness to histamine (PC20). Rijcken et al [1993] showed a similar association with any positive skin test and hyperresponsiveness after adjusting for presence or absence of respiratory symptoms (OR=1.6). In the schoolchildren studied by Sears et al [1991], total serum IgE level was correlated both with symptomatic asthma, but also with bronchial hyperresponsiveness in children who had never experienced asthma, wheezing, eczema or hayfever. Specifically, 5 out of 123 asymptomatic subjects with a sIgE<100 IU/ml were hyperresponsive, and 13 of 75 with a higher sIgE (OR=4.3, 1.5-13.7). These associations persist over time - Dave et al [1990] found that 66% of nonasthmatic atopic subjects in the Natural History of Asthma study were hyperresponsive to methacholine, and that this persisted in 60-70% of subjects for up to eleven years. Only 20% of adults with atopic dermatitis do not suffer from asthma or hayfever, but most of these are hyperresponsive [Barker et al 1991].

Baseline FEV1 is also negatively associated with hyperresponsiveness in many but not all studies. Ulrik [1993] found this association in both allergic and nonallergic asthma, while Carozzi et al [1992], Burrows et al [1992], and Rijcken et al [1993] showed it to be present in general population samples, and Kennedy et al [1990] in nonasthmatic men in five Canadian workplaces. I do not find it surprising that chronic asthmatics, asked to withdraw their medications some hours before undergoing bronchoprovocative challenge, have a diminished baseline FEV1 and increased responsiveness.

Tobacco smoking increases nonspecific bronchial responsiveness [Annesi et al 1988; Sparrow et al 1989; Rijcken et al 1993]. A number of studies have shown an association between eosinophil count and bronchial responsiveness. Rijcken et al [1993] found (log) eosinophil count a significant independent predictor after the effects of FEV1, skin prick tests, report of respiratory symptoms and current smoking were adjusted for. Kauffman et al [1990] reported similar findings in French busdrivers.

The pattern of these associations makes it clear that although increased bronchial responsiveness is a central part of the definition of asthma, nonspecific hyperresponsiveness is certainly not unique to asthma. Specific hyperresponsiveness is much more specific, and this ties asthma to IgE mediated processes (or other types of hypersensitiveness), which may or may not be the case for adult-onset intrinsic asthma.

Descriptive epidemiology of upper respiratory tract allergic disease

Incidence and prevalence in Australia

Hayfever was present in 10.3% of respondents in the Australian National Health Survey 1989-90 [ABS 1991]. Recent symptoms were more likely to be reported in spring and summer. Women suffered more hayfever than men (11.2% versus 9.4%), and the prevalence (long-term and/or symptoms in the previous two weeks) fell off with age from 13.4% of those aged 15-24 years to 6.3% of those aged over 75 years. Only 5% of those reporting hayfever claimed to have experienced it for the first time recently. There were interesting interstate differences, the prevalence being lowest in the Northern Territory (6.2%) and Queensland, and highest in ACT (15.8%), Western Australia and South Australia. As opposed to asthma, where social class differences seem small, hayfever is highest in professional workers (18.2%), clerks (15.5%) and lowest in plant operators/drivers (9.0%) and labourers (9.6%). This may represent a "healthy worker" effect, differential exposure (tolerization), reporting and/or diagnostic bias, or the known urban/rural difference seen in many countries. As will become clear below, a prevalence of 10% for hayfever is far less than the proportion of the population who would be diagnosed as suffering allergic and perennial rhinitis following detailed interview.

Comparison with other countries

The NHANES II [Turkeltaub & Gergen 1991] found a prevalence of allergic rhinitis (doctor diagnosed hayfever, or recent seasonal nasal/eye symptoms) of 9.8% in whites, but labelled a further 20.4% as suffering perennial rhinitis ("frequent nasal and/or eye symptoms that did not vary by both season and pollen during the past twelve months, not counting colds and the flu" and no doctor diagnosis of hayfever). Allergic rhinitis was more common among urban dwellers (10.3%) than rural dwellers (8.3%).

Aberg [1989] reported allergic rhinitis to be present in 4.4% of Swedish conscripts in 1971, and 8.4% in 1981. Allergic rhinitis was more common in urban than in rural recruits, low in those living on farms, and more prevalent in those with higher intelligence scores.

Temporal changes

Emanuel [1988] has proposed that hayfever is a recently developed disease. He could find no systematic reference to it earlier than the sixteenth century, and until the 1830's the only form recognised was the "rose cold". Bostock at this time described "catarrhus aestivus", but was only able to collect 28 cases, while others described it as rare. By 1860, an anonymous writer cited by Salter [1860, p 288] commented on "...the more general and more severe development of hay-fever which appears to have recently occurred". In a similar vein, Blackley wrote in 1873 that "hay-fever has...been considerably on the increase", and was most common among the educated and upper classes, and uncommon among farm workers - a pattern matching that found by Aberg [1989] and by the Australian National Health Survey. By the turn of the century, Hay Fever associations were being formed with thousands of members. More recently, hayfever prevalence has continued to increase, in line with that seen for asthma.

Descriptive epidemiology of atopic dermatitis

Larsen and Hanifin [1992] have reviewed epidemiological studies of atopic dermatitis. Large surveys of children prior to 1960 put the prevalence at 3%. The prevalence of questionnaire diagnosed infantile (prior to age 2 years) eczema in Busselton in 1970 [Turner et al 1974] by contrast was 5.6%, and up to the age at survey (maximum 17 years) 8.8%. In the British National Childhood Development Survey in the same year, the cumulative incidence in 7 year olds was 12.2%. They note use of terms such as "eczema" in questionnaires gives much higher prevalences than surveys that include dermatological examination. The twin studies performed by Larsen et al [1986 1993] and reviewed below, documented similar increases over a ten year period in prevalence in a study where diagnoses were carefully validated. In the 1958 birth cohort of the British National Child Development Study [Williams et al 1994], the prevalence of childhood eczema on physical examination in social class I (highest) was twice that seen in class V (lowest) - reminiscent of the trends seen earlier for hayfever.

Environmental aetiological agents involved in asthma

Allergens

Antigens can vary quite considerably in size and nature. Nevertheless, the majority of allergens are (larger) proteins, and usually contain numerous disulphide bonds [O'Hehir et al 1991]. A number of significant ones have recently been identified as enzymes, especially proteases (eg cat and mouse urinary and house dust mite faecal allergens). Radauer et al [2008] categorized 847 sequenced allergens, and found 119 were hydrolases, and 56 of these peptidases. These molecules retain some activity, which could potentiate inflammation in the lung, and thus sensitisation. Gough et al [2001] have shown that the major HDM allergen Der p I can enhance the IgE response to ovalbumin, an effect abolished by blocking the enzymatic activity of the allergen. Der p I has several actions, cleaving occludin, a protein in the epithelial tight junction, and several cell surface receptors such as CD23, CD25 and CD40 [Furmonaviciene et al 2007]. A number of small molecules such as toluene diisocyanate and plicatic acid, that cause asthma in some exposed individuals, do not cause immediate type (I) hypersensitivity, even though challenge with these agents can lead to a late asthmatic response. The mechanisms here are not yet understood.

Development of sensitisation to novel allergens occurs over varying durations of exposure. Occupational exposure to animals is usually associated with a latency period of months; low molecular weight chemicals, two to four years. Some individuals only become sensitised after many years of exposure to a particular allergen - Schwartz [1952] gives the example of baker's asthma. In the case of ubiquitous agents like house dust mite allergen, most atopic children have detectable levels of anti-HDM IgG before one year of age, although specific IgE will only rise after two to three years.

Several studies have documented a secular increase in allergen sensitisation parallel to that seen for clinical allergic disease. Broadfield et al [2002] suggest that the decrease in positive skin prick test rates seen with increasing age is largely due to this cohort effect, in that repeat testing after a nine year interval in adults showed a slight increase, while strong birth cohort effects were seen on both occasions.

Aeroallergens

Approximately 90% of childhood asthmatics and 80% of adult asthmatics are allergic to aeroallergens, that is they develop acute and/or delayed tissue reactions upon application of these agents to skin, nasal or respiratory mucosae. The effects of these agents on inhalation is to precipitate asthma, as well as increased nonspecific bronchial responsiveness. A number of recent studies have documented the associations between sensitisation to particular allergens and asthma (see Table). Of particular interest is the finding by Ohman et al [1993] that new asthma in 38 over 60 year olds in the Normative Aging Study cohort was preceded by anti-house dust mite specific serum IgE increases (RR÷2), but not total sIgE, anti-cat (save slightly in nonsmokers) nor ragweed IgE.

Table 3. Association between specific allergen sensitisation and asthma and/or bronchial hyperresponsiveness from recent community based studies as adjusted odds ratios from logistic regression.
StudyHouse Dust MiteCatAlternariaRye Grass
Lau et al [2000]4 (3.6-4.1)
Peat et al [1992]7 (3-14)3 (2-6)2 (0.5-3)1 (0.5-2)
Peat & Woolcock [1991]6 (4-10.5)4 (2-8)1 (0.5-3)1 (0.5-2)
Sporik et al [1990]34 (4-742)------
Sears et al [1989]7 (1-13)4 (2-8)1 (0.5-2)1 (0.5-2.5)
Gurgen et al [1992]3 (2-5)0.5 (0-1)5 (3-9)1 (1-2)
Cookson et al [1991]19 (2-154)N.S.*8 (2-38)

* N.S. Not significant in logistic regression - odds ratio not cited.

Pollens: These historically were the first allergens to be associated with allergic disease, mainly by the association between disease and season, and by disease and proximity to plants (eg "rose cold"). A broad variety of pollens are known to be allergens. In the USA and Europe, a major cause of pollinosis is ragweed (Ambrosia artemisifolia). In Australia, by contrast, ragweed is not widely distributed. A number of native and imported trees and grasses are more important. These include high protein European grasses such as timothy and perennial rye grass, a number of wattles and flowering gums.

In the NHANES II study [Gergen et al 1987], sensitisation to pollens is higher in urban areas than rural areas. This can be interpreted either as due to development of immunological tolerance in children brought up in rural areas, or due to adjuvant effects of urban pollution etc. Anecdotally, individuals brought up in rural areas have described development of hayfever for the first time on returning home from a period away.

Fungal spores: Alternaria and Aspergillus spp are the two main fungi associated with asthma. The table above demonstrates that sensitisation to Alternaria is a risk factor for asthma, while O'Hollaren [1991] implicated it as a significant associate of near fatal asthma attacks, and Dales et al [2003] show it underlies "thunderstorm" asthma in Canada (the increase in hospital presentations of asthma during storms; in passing, Melbourne thunderstorm asthma has been suggested to be due to low MW rye grass pollen allergens). Peat et al [1992] found it more important than house dust mite (see below) as a cause of sensitisation in asthmatics in dry inland Australia. Visible mould in houses is associated with reported increased asthma in temperate climate studies [Strachan 1988], though this is also correlated with increased house dust mite allergen [Voorhoost 1969]. Belanger et al [2003] studied 849 infants with a family history of asthma. They found both maternally reported exposure to mould and measured fungal spore counts in the home to predict wheezing in the first 12 months of life, while measured house dust mite allergen levels were not associated. Salo et al [2005] found detectable levels of Alternaria allergen in 95% of US houses in the National Survey of Lead and Allergens in Housing, highest where visible mould or dampness was present.

Castor bean: This is an uncommon exposure, but dust arising from castor bean processing is sufficiently potent to evoke allergic sensitisation even in nonatopic individuals [Kemeny & Diaz-Sanchez 1993]. Ricin (the notorious toxin) present in the castor bean enhances sensitisation to other allergens presented at the same time. It is interesting to speculate if other agents can act in this way, eg cigarette smoke [Venables et al 1989].

House dust mite: It had been known since the 1920s that challenging some asthmatics with dust from their homes could precipitate asthma, but it was not until 1967 that Voorhoost first identified mite products as the allergens present in house dust [Voorhoost et al 1969]. These were subsequently identified as being the mite faeces [Tovey et al 1981].

The bulk of house dust mites (HDM) are pyroglyphid mites, living largely on human and other scales. The two main pathogenic species in Australia are Dermatophagoides pteronyssinus and D. farinae. The Euroglyphus genus, particularly sp. E. maynei, are also common in Australia, and may produce major allergens. A number of other storage mites, that is mites associated with stored grains in silos, homes etc are also known to be associated with allergic disease in some individuals.

HDM biology: HDMs are known to exist in great numbers in soft furnishings and bedding in most Western type homes [Platt-Mills et al 1989]. Their growth requires a relative humidity of 60% or greater (>7 g/kg), and is optimal at temperatures between 17-25 C. In most countries therefore, their numbers exhibit (late) summer peaks and winter/spring minima. This occurs in Queensland [Domrow 1970], where a February peak is observed, while Australia wide, more humid coastal cities tend to harbour greater numbers of HDM than inland towns [Green et al 1986]. Numbers are lower worldwide at higher altitudes for the same reason.

Nature of HDM allergens: The allergens produced by various species of HDM are now well characterised. The genes for a number have been cloned and sequenced. Three main classes I,II and III are recognised (and others classes up to VII). Classes I and III are thiol and serine proteases respectively, and weigh 25-29 kD [Chapman 1990]. Class II allergens are thought to derive from lysozyme (14 kD) [Thomas 1993]. T-cell clones sensitised to particular allergens have now been isolated (the significance of which is reviewed later). Kort and Kniest [1994] have shown these allergens are stable under normal household conditions for up to four years.

Evidence implicating HDM in asthma: A number of environmental studies, longitudinal and intervention studies have conclusively shown a role for house dust mite exposure in increasing severity of symptoms of asthma [Voorhoost et al 1969; Korsgaard 1983; Pollart et al 1989], as well as atopic dermatitis [Sanda et al 1992]. Classically, one treatment for asthma has been to move to high altitude or dry climates - both situations where HDM numbers diminish and also asthma prevalence. By contrast, asthma tends to be more prevalent in climates where year round humidity is high, and many asthmatics from dry inland regions note a deterioration in condition on approaching the coast [eg Ordman 1968]. Observational studies such as Woolcock's work in Papua-New Guinea have suggested that increasing HDM numbers might also lead to an increased prevalence of disease. Here, [Dowse et al 1985], the introduction of blankets (and thus HDM) has coincided with a rise in asthma prevalence from 0.3% to 7.3% in some areas, which is not seen in other areas where mite counts have not risen [Turner et al 1988]. In studies of adult asthmatics seen in Brisbane practice, 85% of young adults [Wright & Derrick 1975] and 65% of hospital patients were sensitised to HDM [Clarke & Aldons 1979].

A well known study is that of Sporik, Holgate et al [1990], who performed an 11 year concurrent cohort study of children with a family history of atopy. There were 67 out of an original 93 children followed from 1978 to 1989. Importantly, dust from the child's mattress, bedroom and living room floors were collected in 1979-80 in 59 cases, as well as in 1989. Subjects completed a questionnaire in 1989, underwent bronchial provocation using the protocol of Yan et al [1983] on four occasions, as well as SPT, RAST and total sIgE level determination.

There were 35 skin-atopic and 32 non-skin-atopic subjects. Bronchial responsiveness was present in 23, and asthma (BHR plus wheeze in the last year) in 17. Of the asthmatics, 16 were skin test positive to house dust mite. Forty percent of those SPT positive for HDM at age 11 had also demonstrated sensitisation at age 5.

There was a marginal association between infant exposure to high levels of HDM and level of skin sensitisation. Early exposure was associated with later wheeze, and age at first episode of wheeze, the latter relationship being strongest in skin-atopic children (r=-0.66, P=0.001). This last point suggests that these genetically predisposed individuals avoided asthma if they were brought up in an environment low in HDM.

Other studies have not replicated this effect. For example, Burr et al [1993] followed 453 children with a family history of atopy from birth to age 7 years. The coefficient of variation for repeated determinations of Der p I (Class I) allergen in a mattress was 15%. One third of the children were diagnosed as asthmatic (by the study paediatrician). Neither wheeze with atopy nor sensitisation to HDM on skin testing at age 7 were predicted by level of allergen measured in the first year of life.

Lau et al [2000] also failed to find any variation in risk versus environmental levels of mite allergen. A sample of 1314 infants born in 1990 were followed longitudinally by the German Multicentre Allergy Study, 939 to age 7 years. One-third of the cohort were selected to have two or more atopic relatives or an elevated cord sIgE.

Children with incomplete followup did not differ from those with complete data save that their parents were more likely to smoke and were less well educated. Doctor diagnosed asthma was found in 6.1% at age 7 years, and 10% reported wheeze in previous twelve months. There was a strong association between sensitisation to HDM (and cat) allergens and wheeze, asthma and bronchial hyperresponsiveness to histamine. While higher levels of environmental HDM allergen at age six months predicted HDM sensitisation (at age 3 years, OR=8.8 comparing lower and upper quartiles of exposure), it was not correlated with risk of wheezing at age 7, age at onset of wheezing, BHR or asthma.

An explanation for this and other negative findings [eg Sporik et al 1993; Pauli et al 1993] of a dose-response effect is that possibly all individuals in the study were exposed to a high enough level of allergen to precipitate disease. An international workshop [Platt-Mills et al 1989] has suggested that a level of 2 ug of Der p I per gram of dust (roughly equivalent to 100 mites per gram of dust) might represent a threshold above which sensitisation is likely to occur, and a level 5 times this as a level likely to precipitate symptoms in sensitised asthmatics. In homes in Lismore, Australia, the mean level of Der p I levels was 83 ug/g, and in inland towns, 11 ug/g [Peat et al 1993]; in Strasberg, Germany, mean levels were 24 ug/g [Pauli et al 1993]; in homes of controls in Dorset, UK, 17 ug/g [Sporik et al 1993]; in 252 homes of asthmatics from around the US, 95% contained one sample containing over 100 mites/g, and one third had an average count (multiple sites within the house) greater than 500 mites/g [Arlian et al 1992]. Arguing against this explanation for the results of Lau et al [2000], levels were lower (median level in mattresses 5.6 ug/g), so that one third of the sample fell below the 2 ug/g threshold.

The role of changes in housing and cleaning practices that promote HDM growth, such as central heating in temperate climates, "tighter" buildings with lower natural ventilation, cold water washing, and vacuuming rather than (airing and) beating carpets, has been advanced as one hypothesis explaining the rise in asthma prevalence seen this century, and especially this last twenty to thirty years. Since Australia has one of the highest rates of wall-to-wall carpeting in the developed world [Tovey, personal communication], household factors might be important in the high prevalence of asthma here.

Cockroach: Increasing interest has been developing about the role of cockroach allergens in asthma and allergy. The presence of anti- cockroach IgE has been shown to be a risk factor for acute asthma in a number of recent studies. It tends to be most important in asthmatics living in poor inner city accommodation. In Gelber et al's case-control study of acute asthma admissions [1993], presence of anti-Blattela IgE (on RAST) was a significant risk factor (OR=5.0, 1.4-17.8). Allergy to cockroach has been reported to be associated with more severe disease [Kang et al 1992].

Animal danders: Hypersensitivity in proximity to animals has also been known for a long time. A variety of animals can lead to sensitisation. Cats produce three main allergens. Fel d 1 is a protease, and is found on the fur, within sebaceous glands, and in saliva. Fel d 1 survives well in the domestic environment, and has been detected in homes 10 years after a cat was last present. Allergen particles are relatively large (30 æm), and so tend to settle fast, but cause immediate symptoms on disturbance. In Platt- Mill's study, anti-Fel d I IgE was significantly associated with asthma (OR=4.7, 1.5-14.4). Allergy to cats is associated with asthma in Australian [Peat et al 1991] and New Zealand children [Sears et al 1989], but was not in a study of US adults [Gergen et al 1992]. Demonstration of an association between exposure to cats and the development of asthma (again as opposed to exacerbation of existing asthma) has been complicated by the fact that sensitised individuals often have removed cats from their homes [Brunekreef et al 1992].

Rats and mice are associated especially with sensitisation in laboratory workers who are involved in handling these animals. In mice, the predominant allergen is a urinary protein Mus m I, possibly a lipid transporter protein [O'Halloren 1992].

Low molecular weight chemicals

A number of low molecular weight chemicals used in modern industry are potent inducers of asthma. A few of these act as haptens, eliciting production of IgE to complexes of the chemical with serum albumin or other proteins. Examples of this are the anhydrides such as phthalic and trimellitic anhydride. By contrast, in individuals sensitised to toluene isocyanate, usually fewer than 15% elaborate significant levels of circulating IgE [Zeiss 1991]. Cartier et al [1989] found elevated anti-isocyanate IgG in 21 of 29 sensitised workers, but IgE in only 9 workers. Isocyanates may also act directly on cAMP release and/or the þ-adrenoceptor to cause asthma [Montanaro 1992].

Ingested allergens

Infant allergic disease is often associated with sensitisation to particular foods, and delaying the onset of solid food in the diet to after six months is associated with a fourfold reduction in the risk of developing infantile eczema. Avoidance of allergens such as cows milk in the first year of life does not protect against the subsequent development of asthma [Burr et al 1993].

Viral respiratory tract infection (VRTI)

A number of exacerbations of existing asthma in children and less so in adults are associated with intercurrent VRTI.

Respiratory Syncitial Virus (RSV)

Infections with this virus lead to wheezing in children, especially under the age of two years - bronchiolitis. Numerous studies have documented increased rates of subsequent asthma in children who have suffered from bronchiolitis. However, one risk factor for the development of bronchiolitis is a family history of asthma, suggesting infection is not an inducer of asthma save in those already at increased risk. RSV infection is associated acutely with increases in bronchial responsiveness, and small airway obstruction. Welliver [1981, 1986] has shown that anti-RSV IgE is produced by individuals with RSV infection that has caused acute wheezing.

A recently identified similar virus, human metapneumovirus [Van Den Hoogen et al 2001], seems to give rise to a similar spectrum of disease.

Adenovirus

Adenoviral infection is the second (though far less) common cause of bronchiolitis after RSV. Hogg [1992] has presented evidence suggesting that chronic (or latent) infection might underlie persistent asthma. His group, using PCR to amplify the genes encoding the "immediate early" proteins (E1A), demonstrated higher copy numbers of this gene in the lungs of patients with obstructive lung disease compared to controls. One of the E1A proteins sensitises infected cells to lysis to TNF-alpha - an agent suggested to be important in the allergic late phase reaction. Influenza and parainfluenza infections are also associated with wheezing in some children and with the production of antiviral IgE.

Bacterial respiratory tract infection

Historically, hypersensitivity to bacterial antigens was supposed to underlie the association between asthma and respiratory tract infection. In the case of intrinsic asthma, Rackemann hypothesised that sensitisation to bacteria was the causative factor [Swineford 1968]. There is little evidence currently to support this hypothesis, though in atopic dermatitis, where staphylococcal infection of lesions is common, allergy to enterotoxin (specific IgE and increased basophil releasability) has been documented [Leung et al 1993].

Chlamydial bronchitis/pneumonia

Chlamydia pneumoniae has been reported to be a significant cause of community acquired respiratory tract infection, and circulating specific IgG against this pathogen is found in approximately half the adult population. In one prospective US study [Hahn et al 1991], C pneumoniae was the cause of 3 out of 11 (10%) atypical pneumonias, and 16 of 338 (5%) episodes of bronchitis. These individuals were more likely to have presented with wheezing (OR=2.1, 95%CI 1.1-4.2). A nested matched case-control analysis of seropositivity to C pneumoniae (N=71 pairs) found an odds ratio of 7.2 (95% CI 2.2-23.4) for development of wheezing during or in the six months following the acute illness.

Several replications of this association have been published. Hahn and McDonald [1998] examined a further 163 individuals presenting with acute wheezing or chronic asthma and found 20 with serology consistent with acute infection. Of these, ten were wheezing for the first time, and five developed subsequent persistent asthma. Von Hertzen et al [1999] found a 3-4 fold increase in the prevalence of IgG against C pneumoniae among 332 Finnish asthmatics compared 98 controls. Cook et al [1998] also found 16 of 46 chronic asthmatics seropositive compared to 12.7% of 1518 hospital controls, but 14.6% of 123 admissions with acute asthma.

Hahn and Peeling [2008] studied anti-hsp60 antibody titres in 86 asthmatics and 52 nonasthmatic controls. The asthmatics were three times more likely to be sensitized to the C. pneumonia hsp60 (heat shock protein 60) than the controls, and seropositivity was associated with lower FEV1. This replicates the earlier work of Huittinen et al [2001] (24 cases, 62 controls). The C. trachomatis hsp60 is believed to underlie chronic effects of infection with that organism (eg trachoma, salpingitis), via molecular mimicry and subsequent development of autoantibodies.

Bronchiectasis

This condition has been noted to be associated with asthma in studies from South East Asia, notably Hong Kong, where the incidence of bronchiectasis remains much higher than the Western world. If chronic exposure to intrapulmonary bacterial allergens was important in the aetiology of asthma, bronchiectatics would be expected to be at risk. I am not convinced that there is no mislabelling of symptoms in these studies; BHR is increased in these patients, but does not prove asthma per se.

Bacterial sinusitis

There is undoubtedly a strong association between asthma, especially attacks requiring hospital admission, and sinusitis [Spector 1978; Slavin 1982; Zimmerman et al 1987]. X-rays reveal sinus opacification or mucosal swelling in up to 65% of such patients, though one study by Zimmerman et al [1987] found no correlation between radiological changes and severity of asthma. Spector [1978] and Slavin [1982] have argued that since bacterial culture of chronically inflamed sinuses is positive in between 20-80% of cases, and that some older studies suggest bacterial sinusitis can lead to seeding of the lungs, antibiotic therapy in severe asthma might lead to an improvement in asthma. They and other respiratory physicians feel that patients benefit from clearance of the sinusitis, a claim given support by some intervention studies. Slavin [1982] reported a series of 33 asthmatics, most of whom underwent intranasal sphenoethmoidectomy. Of these patients, 85% reported subjective improvement, 10 patients were able to stop oral steroids while a further 15 out of 18 reduced their dose from 20 mg/d of prednisone to 3 mg/d.

Another possible mechanism for a causative link between upper respiratory disease and asthma suggested by Slavin is the association between nasal mucosal irritation and reflex bronchoconstriction. Brugman et al [1993] in a rabbit model have shown that induced sterile sinusitis (using recombinant C5a sinus lavage) leads to increased bronchial responsiveness to histamine. This seemed to be mediated by post-nasal drip (of mediators), as altering drainage by changing head posture abolished it.

The hygiene hypothesis

Strachan [1989] was the first author to report a correlation between birth order of a child and their risk of atopic disease. His inference from the rates of hayfever in the 1958 UK National Birth Cohort, was that later born children might be exposed to pathogens brought home by other siblings, pathogens that might channel the developing immune system away from the atopic diathesis.

Sibship size and birth order

It has been consistently confirmed that being a member of a larger family is protective against atopy. Karmaus and Botezon [2002] is a metaanalysis of 53 studies. The weighted average odds ratio for atopic dermatitis for children with three or more siblings was 0.66, for hayfever was 0.56, and for asthma 0.93 (although 0.72 excluding the study of 148000 Swedish military recruits of Braback [1999]). Skin prick test reactivity followed this trend (OR=0.62). The effect is strongest for those late in the birth order.

McKeever et al [2001] present a birth cohort from the West Midlands GP Research Database of 29000 children. Atopic dermatitis and hayfever were descreased for thise with three or more older siblings (Hazard Ratio=0.7), but risk of asthma was decreased only for disease with onset after two years old (HR=0.7); for earlier onsets, the risk was increased (HR=1.4): it is possible the latter represent wheezy bronchitis, with a more benign prognosis. Interestingly, they found atopic disease was less common among twins and higher multiples, compared to singletons.

Farm living

A large number of studies have looked at the previously mentioned rural-urban differences in asthma prevalence and attempted to look for the mechanism involved. Kilpelainen et al [2000], for example, showed that the farm environment reduced risk of asthma in young adults either with or without a family history of atopy. Riedler et al [2001] found exposure to stables and/or farm milk was protective only if prior to one year of age, and more so if the mother had been exposed to one or both of these factors during the pregnancy. Studies of children brought up in the anthroposophist lifestyle also show reduced risk of atopy, but this could be due to a variety of factors (including breast feeding, non-immunization and consumption of fermented milk products).

Viral infections associated with poor hygiene

Matricardi et al [2000] describe a case-control study of the relationship between atopy and seropositivity showing prior exposure and/or infection for Hepatitis A, toxoplasmosis and H. pyloridis. The 240 controls were twice as likely to have detectable levels of anti-HAV IgG circulating than the 240 cases. The authors regarded HAV infection as a proxy for poor hygiene. This was not found by Bodner et al [2000].

Gastro-oesophageal reflux (GOR)

One of the earlier theories about the pathogenesis of asthma suggested that the disease was due to occult (nocturnal) aspiration of gastric contents. A large number of studies have shown that GOR has a high incidence in asthmatics, especially children with nocturnal wheeze, and that antireflux therapy can lead to a clinical improvement in severity of asthma. More recent theories have invoked vagally mediated bronchoconstriction due to distal oesophageal irritation [Mansfield et al 1981] rather than a direct attack of acid on the airways.

The endpoints in many of the studies in this area seem rather soft, and a controlled trial is necessary. In addition, it is accepted that theophylline, widely used in the US, can lead to decreased lower oesophageal sphincter pressure and thus reflux. This might partly explain an association between asthma and GOR. A final point is that the definition of "pathological" reflux is blurry. In studies where the acid reflux test is used alone, the false positive rate can be high.

Lack of breast feeding

A large number of studies have found that breast feeding protects to some extent against the development of atopy, most obviously in the case of infantile atopic dermatitis and food allergy. The effect is small, and so not detected in all studies [Duffy & Mitchell 1993; Martinez et al 1992; Burr et al 1993; Arshad et al 1993], and is confounded by other factors associated with the maternal choice to breast feed (social class, cigarette smoking). For example, in 1022 infants studied by Martinez et al [1992], wheezing before one year of age was predicted by maternal age, smoking, birth order but not breast feeding. Sears et al [2002] found breast feeding to be associated with a two-fold increased risk of subsequent atopic sensitisation and asthma in a New Zealand birth cohort born 1972-3. Those authors review the literature and suggest that the protective effect is only on disease in early life, while risk after six years of age may be increased.

Dietary probiotics/lactobacillus

Several randomized controlled trials have examined whether ingestion of milk products containing lactobacillus culture may influence risk of atopic disease. Kalliomaki et al [2001] randomized high-risk mothers (one or more atopic relatives) and their infants to Lactobacillus GG or placebo. Atopic eczema was halved in the probiotic group (15/64 vs 31/68). A similar result [Isolauri et al 2000] was obtained in a smaller trial on infants already suffering from eczema, with a drop in severity score from 13 to 1 (P=0.002).

Low Birth Weight

Low birth weight (LBW) is a risk factor for both neonatal and long term respiratory disease. It had been usually accepted as a cause of childhood obstructive airways disease, and this had been attributed to the increased incidence of need for ventilatory assistance, hyaline membrane disease, or meconium aspiration during delivery. For example, lower respiratory disease is present in 8% of very LBW infants at 40 weeks of age [McCormick 1985]. A number of long term follow up studies of ventilated infants have found chronic airflow limitation in up to 70% in adulthood [Northway et al 1990].

LBW children, even in the absence of perinatal respiratory complications, also suffer from increased numbers of respiratory tract infection in the first eighteen months of life [Drillien et al 1958]. It is known that lower respiratory tract infection before age 2 years is a predictor of both diminished lung function and chronic respiratory symptoms [Gold et al 1989]. Martinez et al [1988] showed that wheezing in the first year of life was associated with a prior increased total respiratory resistance, but did not investigate antenatal and perinatal events that may have led to diminished lung function. Schwartz et al [1990] report that LBW is also associated with childhood asthma and wheeze (OR=1.4), as did Arshad et al [1993] in a cohort study up to age two years (OR=3.0, 95%CI=1.4-6.5). Arshad et al did not detect an effect of prematurity, but Rona et al [1993] showed that LBW was not a risk of wheeze after adjustment for prematurity. Prematurity was a significant risk factor, with the odds of frequent wheeze rising 1.1 (95% CI=1.0-1.2) for every one week of prematurity. Evans et al [1998] found the strongest predictor of asthma between 2--5 years of age was radiographic evidence of bronchopulmonary dysplasia in a cohort of children with birth weight <1501 g (mean gestational age at birth 30 weeks). This effect on asthma is small in terms of attributable proportion, and so has not been detected in smaller longitudinal studies [eg Leeder et al 1976].

A possible confounding factor is that asthma in the mother may predispose to premature labour. Kramer et al [1995] found such a relationship (OR=2.3, 95%CI=1.1--4.6). Evans et al [1998] found a similar, marginally significant, trend (OR=1.9, 1.0--3.5) for asthma arising at ages two to five years.

Other prenatal/perinatal influences

A number of studies have shown that umbilical cord total sIgE level predicts the risk of later allergic diseases. Levels are extremely low at birth - Halonen et al [1991] report a geometric mean of 0.09 IU/ml. In that study of 1074 infants, those that developed eczema in the following nine months had a significantly higher mean cord IgE of 0.16 (P=0.002). There was tracking of sIgE levels with a correlation of 0.44 between sIgE level at nine months and at birth. Kjellman and Croner [1984] used a cutoff of 0.09 IU/ml to predict allergic disease developing by age six years in 1651 children. Only 12% of their sample exceeded this threshold. For asthma, the odds ratio was 7.8 (95%CI=4.7- 13.1), and for atopic dermatitis, OR=5.5 (4.0-7.7). These correspond to sensitivities of 0.49 and 0.36 respectively, so this was not a useful screening test.

Such effects must either be genetic, and/or due to intrauterine influences. There is evidence to support both mechanisms. For example, Halonen et al [1991] report a seasonal variation in cord sIgE with a spring- summer peak, a pattern coinciding with that seen for allergic diseases as well [eg Duffy & Mitchell 1991; Arshad et al 1993]. This seasonal effect (which is not usually large) implicates pre/perinatal events in the genesis of asthma. Maternal diet has been correlated with the specificity of the cord IgE [Michel et al 1986]. Such maternal effects complicate the genetic modelling of atopy.

Handedness

Several studies have now claimed an association between sinistrality and atopic conditions, including asthma. Norman Geschwind first suggested an association [Geschwind & Behan 1982] might exist, basing this on a theoretical mechanism for embryological lateralisation. All replications to date have been performed by psychologists rather than epidemiologists, in small samples [Benbow 1986; Pennington 1987]. One intriguing study [Weinstein et al 1990] has suggested increased sinistrality among children of asthmatics. A possible mechanism for all this is via low birth weight [Williams et al 1992].

Psychological mechanisms

Exacerbations of asthma have been associated with emotional stress since antiquity, and the psychoanalysts such as Alexander developed psychosomatic (metaphorical) explanations for the aetiology of asthma. A entire issue of Psychological Medicine [1990] reviewed the literature, and concluded that psychological mechanisms other than those observed in any group of chronic disease sufferers were not important [eg Rutter 1970]. That is, denial of symptoms in severe asthmatics is a risk factor for undertreatment and thus sudden death, and behavioural and emotional disorders can arise as a consequence of severe illness.

Tobacco smoke

The existence of a relationship between cigarette smoking, either active or passive, and the development of asthma remains a little unclear. There seems no doubt that smoking does lead to COPD, and passive smoking can be demonstrated to diminished lung function, especially on measures of small airway function [Kauffman et al 1989; Duffy & Mitchell 1993; Rona & Chinn 1993]. But, though a number of studies of children of smokers have suggested an increase in asthma (and acute respiratory illness), other large cross-sectional studies have found no association. Part of the reason for this is a reciprocal interaction, in that asthmatics seem to be less likely to take up smoking. In view of this, The UK Scientific Committee on Tobacco and Health concluded that passive smoking was more likely to be exacerbating preexistent asthma rather than causing disease, and the California EPA the converse [Infante-Rivard et al 1999].

A recent paper by Hjern et al [2001] describes results from the Swedish Survey of Living Conditions (11381 subjects). Both smokers and the children of smokers had decreased rates of asthma and other allergic disease compared to never-smokers. In view of the papers described below, the authors queried the direction of causation of this finding.

Weiss et al [1985] first reported a two-fold increase in atopy among offspring of smoking parents. This was replicated by Martinez et al [1988; Ronchetti et al 1990], who demonstrated a concurrent rise in BR, sIgE level and eosinophilia in children of smoking parents. A smaller effect on risk of persistent wheeze was found in the Six Cities study [Gold et al 1993] in US offspring of smokers (OR=1.3, 95%CI=1.1-1.7). Maternal smoking was found to predict asthma up to two years of age (OR=2.2, 95%CI=1.5-3.4) by Arshad et al [1993], but not by Burr et al [1993] in a similar cohort study from birth to seven years of age. Duffy and Mitchell [1993] could not demonstrate any increase in reported wheeze in such children, although a slightly diminished FEF25-75 did differentiate offspring of smokers from those of nonsmokers. Strachan et al [1990b] could not demonstrate a relationship between salivary cotinine and wheezing in 770 seven year olds. Lebowitz [1977] found an association (RR=2.2 for "ever" versus "never" smoked in males) between smoking habits and asthma in non-skin-atopic adults from the Tucson Epidemiologic Study of Obstructive Lung Disease cohort study. Heavy smokers were only 1.5 times as likely as never smokers to report asthma, and if asthma unaccompanied by other diagnoses (chronic bronchitis and emphysema) was used, no relationship was found. Vesterinen et al [1989] found smokers to be twice as likely to develop asthma as nonsmokers in their large prospective study. Tashkin et al [1993] reassuringly report that marijuana does not seem to increase bronchial responsiveness in the way that tobacco smoking does.

Air Pollution

Air pollutants can be broadly broken into photochemical oxidants (ozone and NOx), sulphur dioxide and SO2-particulate complexes, acid aerosols (such as sulphuric and nitric acid) and particulates. While air pollution is associated with chronic bronchitis, and exacerbates existing asthma, a causative role in the occurrence of asthma is not well established. Ozone has been shown to cause airway inflammation that is associated with increased nonspecific bronchial hyperresponsiveness.

Epidemiological evidence linking asthma and allergy with air pollution is less strong than that for chronic bronchitis. Whincup et al [1993] have pointed out that the increases in the prevalence of asthma seen in the UK run counter to the decreases in rates of smoking and exposure to particulate and SO2 pollution in that country. In a similar vein, von Mutius et al [1992] found bronchial hyperresponsiveness, asthma and allergic rhinitis less common (though bronchitis more common) among schoolchildren in heavily polluted Leipzig (Eastern Germany) when compared to less polluted Munich (Western Germany). Corbo et al [1993] also found reported asthma no more common in Rome and Civitavecchia (high pollution - prevalence 7.8%) than in rural Viterbo (low pollution - 5.6%, OR=1.3, 95%CI=0.9-1.9), and skin test positivity rates were the same in each region (21% positive to any of eight allergens).

SO2 and particulates both arise from the burning of fossil fuels. Dense particulate smogs such as that affecting London in 1952 are associated with marked mortality and morbidity due to respiratory disease, and Schwartz [1992] has presented longitudinal data from six US cities to show that total mortality is linearly related to particulate levels (TSP) with no threshold of effect. Studies have suggested an association between daily hospital attendance rates for asthma [Schwartz et al 1993], asthma medication use [Pope et al 1991] and particulate levels (PM10), while limited animal work has suggested that hydrocarbon particles can act as adjuvants for allergens.

Low concentrations of inhaled SO2 do precipitate episodes of bronchoconstriction in asthmatics - notably the dose inhaled while consuming food or drinks containing sodium metabisulphite. In animal models SO2 facilitates allergic sensitisation [Reidel et al 1988]. Studies have suggested a relationship between SO2 and hospital admissions for asthma [Tseng & Li 1990].

Dietary Sodium

Burney [1987] has reviewed epidemiological evidence that implicates the high sodium diets consumed by humans in developed societies in increasing bronchial responsiveness. He argues, correctly, that the rise in the prevalence of asthma in African and Polynesian peoples moving from traditional to Western lifestyles must have an environmental basis. He has found that within England, regional asthma mortality is correlated with mean table salt purchases, and that in one survey of nonspecific bronchial responsiveness, 24 hour urinary sodium excretion was correlated with PD20 to histamine, after adjusting for skin atopy [Burney et al 1986]. Pistelli et al [1993] more recently reported that personal table salt use was associated with wheezing in a sample of 2439 adolescents (for wheeze outside of a cold OR=2.2, 95%CI=1.3-3.8; for exercise induced wheeze OR=2.2, 95%CI=1.3-3.4). Urinary potassium but not urinary sodium excretion was associated with BHR. Demissie et al [1996] performed a case-control study of childhood asthma (187 cases, 145 controls) versus salt intake measured by food frequency questionnaire. Although there was no association with previously diagnosed asthma, BHR to methacholine increased linearly with salt intake. Ehrlich et al [1996] failed to find an association with clinical asthma in a similar South African study (368 cases, 294 controls).

Several trials of a low or high sodium diet in asthmatics have been performed. Javaid et al [1988] found that doubling salt intake increased bronchial responsiveness to histamine in nine out of ten asthmatics in one dietary challenge study. This finding was replicated by Carey et al [1993] in a double blind crossover trial of sodium supplementation in 22 asthmatics.

Dietary Selenium

Selenium levels in New Zealand soil are very low, and although human disease had not been linked to deficiencies in that country, some domestic animals have required supplementation. Since asthma incidence is so high in New Zealand, this was examined as a possible risk factor [Flatt et al 1990]. Serum levels were decreased in asthmatics compared to controls. A number of previous studies have suggested alterations in the selenium-dependent enzyme glutathione peroxidase in asthmatics [Hasselmark et al 1990; Bibi et al 1988]. Hasselmark et al [1993] performed a double blind trial of selenium supplementation in 24 intrinsic asthmatics. A clinical improvement (based on patient report and blind review of results by the investigators) was noted in 6 of the 12 subjects who received the fourteen weeks of supplements, and in only one of the controls. No significant changes were detectable in spirometry or histamine challenge results however. Schwartz and Weiss [1990] reported data from the second National Health and Nutrition Examination Survey, where no association between recent wheeze and selenium intake or level was found.

A study of Auckland infants [Dolamore et al 1992] suggested that non breast fed infants had lower levels of selenium and glutathione peroxidase levels - this could possibly provide yet another explanation of the protective effect of breast feeding on asthma.

Other nutrients and asthma

A number of other cofactors important as antioxidants have been examined for an association with asthma and obstructive airway disease. During the NHANES II [Schwartz & Weiss 1990], candidate nutrients such as dietary beta-carotene, ascorbic acid, copper, zinc and selenium were measured in nearly 7000 30-75 year olds, and included in logistic regression to predict recent wheeze (previous twelve months) and bronchitis. The prevalence of wheeze ranged from 5% of white males aged 30-39 years up to 16% of those aged 70-75 years, and from 7% to 10% of white females. Rates in blacks were a little higher. Wheeze was found to be (independently) associated with total energy intake, niacin and ascorbate levels, the zinc:copper ratio, smoking, family income, and state of residence. I am unsure how much of wheezing in this sample is due to asthma and chronic obstructive disease - the rise in prevalence with age suggested the latter is important. Decreased serum retinol has been found to predict later obstructive airway disease (excluding asthmatics) in other studies [Morabia et al 1989, 1990] - but this was not the case for the NHANES data.

Fogarty et al [2000] found increasing vitamin E intake to be associated with decreasing sIgE level in their Nottingham cohort (N=2633), after adjusting for sex, age and cigarette use. Although vitamin E and vitamin C intake were strongly correlated, any relationship with vitamin C disappeared once vitamin E was entered into the regression.

A number of small clinical trials have evaluated the effectiveness of ascorbate in asthma and found small benefits. Those evaluating fish oils (the omega-unsaturated fatty acids), which have shown a benefit in other inflammatory diseases (notably rheumatoid arthritis), have found either no effect [Arm et al 1988], or a worsening [Picado et al 1988]. No benefit of fish in the diet was found in NHANES [Schwartz & Weiss 1990].

Asthma therapy

This remains a controversial area. The increase in asthma mortality seen in 1969 across several countries was at the time put down to the unregulated use of the nonselective beta-agonist isoprenaline, and more recently, another beta-agonist, fenoterol, has been similarly implicated. Page [1993] has put forward the radical theory that bronchodilator therapy allows increased doses of allergen to reach the lower airway, as well as suppressing immunomodulatory functions of mast cells. These mechanisms might turn an otherwise transient period of wheezing in the life of a child into chronic allergic asthma. Animal models support the idea that this could occur [Page 1991].

Several studies have now shown that regular use of beta-agonist therapy decreases their effectiveness as pretreatment in preventing BHR to histamine inhalation. This does not seem to be the case for intermittent (prn) usage. Swystun et al [2000] found that ten days of regular salbutamol inhalation lead to increased BHR on allergen challenge of 14 mild-moderate atopic asthmatics, both on the early (acute drop in FEV1) and late (7 hours post challenge) responses. The latter was of course associated with increased sputum eosinophilia. The authors (including Donald Cockcroft) inferred that mast cell degranulation may be enhanced by chronic beta-adrenergic activation.

Allergy, other diseases and evolution

Since asthma and atopy are such prevalent traits, population geneticists have speculated about mechanisms that have led to genetic susceptibility being so common. The classic model for selection pressure maintaining a common disease is sickle cell anaemia and the heterozygote advantage against malaria that the sickling gene invests [Vogel & Motulsky 1986].

In the study of Sakaguchi et al [1999], Cryptomeria japonica induced pollinosis (allergic conjunctivitis) was present in 8% of a troop of macaques. Specific IgE to Cry j 1 and Cry j 2 was present in 16% of a random sample of 276 monkeys from multiple troops [Hashimoto et al 1994]. The rates of detectable IgE in humans from urban Japan is roughly twice this level, and the rates of pollinosis among Japanese university students 12%. I find it suggestive that the rates are not greatly different.

Parasitic disease

Since the main role of IgE is thought to be protection against parasitic infection, it has been suggested that individuals who mount an excessive IgE response, ie allergic disease, will be less susceptible to parasitosis [Masters & Barret-Connor 1985]. Unfortunately, some anecdotal evidence has been advanced to suggest the opposite, that intestinal or skin (sarcoptes) parasitosis might protect against asthma, while others have concluded that parasitic disease causes asthma. In the case of a protective effect of scabies, however, this might represent allergen hyposensitisation as there is cross reactivity between the antigenic determinants of the house dust mite and Sarcoptes mite.

One of the better known studies examining these hypotheses [Grove & Forbes 1975] was performed in Papua-New Guinea and found a strong negative association between presence of asthma or atopy in the individual and hookworm egg load (all were infected). An earlier Canadian study [Tullis 1970] suggesting a strong positive association was strongly criticised. A number of other case-control studies have found no difference in infestation rates [Masters & Barret-Connor 1985]. Godfrey [1975] noted an inverse association between sIgE level (interpreted as a marker of parasitosis) and prevalence of asthma in The Gambia comparing rural to urban communities, and Gerrard et al [1975] made a similar finding for Metis (Cree Indian descent) and white Canadians (British descent). These findings have been interpreted in population genetic terms, suggesting that a selection advantage (with respect to parasitosis) may have maintained the level of allergic individuals in the population until quite recent times.

IgE and cancer

A number of epidemiological studies have suggested that allergic disease is negatively associated with a variety of cancers. If atopy is protective against some or all cancers, this might represent a mechanism maintaining the high frequency of allergy in the population.

Vena et al [1985] lists sixteen studies [including Cockcroft et al 1979] in addition to their own large case-control study that have examined this question of which 11 showed a protective effect of allergic disease on cancer. Vena et al studied 13 665 cases and 4 079 non-neoplastic controls and found a decreased odds of developing lung cancer in males, and breast and cervical cancer in females who reported previously suffering urticaria. All these and urinary tract cancer in males and digestive tract cancers in both sexes were diminished in those reporting "other allergies", not specified in the questionnaire. A significant adjusted odds ratio of 1.6 was noted for lung cancers in male asthmatics but not in females, a finding supported by other studies - Alameda County [Reynolds & Kaplan 1987]; WW2 veterans [Robinette & Fraumeni 1978]. Overall, Vena et al found a net reduction in all- cancer incidence in allergic individuals.

More recently, results from the Adventist Health study, a cohort study of 34000 subjects [Mills et al 1992], have suggested that a history of reactions to medication (and poisonous plants) may be associated with cancer risk, increasing it in males, and decreasing it in females. Asthma and hayfever were not significantly associated with any particular site of cancer (for asthma and lung cancer - only 62 events - RR=1.2, 95%CI=0.4-3.3). A mechanism suggested for a protective effect by the authors of this study is that asthmatic and allergic individuals are more likely to attend their doctors more frequently - adjustment for time since last visit to doctor attenuated many crude associations in this data set.

Wang and Diepgen [2005] carried out meta-analysis of the association between atopy and a variety of cancers, confirming the increased risk of lung cancer, and concluding a protective effect against several tumour types, notably leukemia, glioma and colorectal cancer. Gandini et al [2005] performed a meta-analysis of 14 studies solely of the association with pancreatic cancer, and found a net protective effect of atopy, but not of asthma.

Diabetes and asthma

The oldest reported association between asthma and another disease (except for that with tuberculosis) is a lower than expected number of individuals suffering from both asthma and juvenile onset diabetes mellitus [Helander 1958]. This was not replicated in one smaller recent study [Stromberg 1995], but Douek et al [1999] compared rates of asthma from the Oxford component of ISAAC to that seen in 181 families of diabetic probands using the ISAAC questionnaire. These authors found wheeze in the previous twelve months to be decreased in the diabetic children (OR=0.36), and to a nonsignificant degree in their siblings. A mechanism for this is that Type I diabetes is seen as a predominantly Th1 lymphocyte mediated process.

Tuberculosis and atopy

There is an extensive older literature on the relationship between asthma and pulmonary tuberculosis, which usually describes a negative correlation. The current immunological paradigm of Th2/Th1 skewing has lead to a re-examination of the relationship, in that a successful immune response to TB is also strongly Th1 (partial or complete loss of the gamma interferon receptor IFNGR1 is associated with disseminated TB), and would be expected to decrease the pool of TH cells available to mount an allergic response. In animal models of asthma, TB infection, and BCG instillation or infection will acutely block the development of allergen sensitization and airway inflammation.

Shirakawa et al [1996] found a negative correlation between tuberculin response (DTH) and presence of allergic disease (including asthma) in Japanese children immunized with BCG, as did Yoneyama et al [2000] in a smaller study, and Aaby et al [2000] in Guinea-Bissau. This was not replicated in Scandinavian studies of children vaccinated in infancy [Alm et al 1997; Strannegard et al 1998] or in young adults vaccinated at age 14 years [Omenaas et al 2000]. In the Scandinavian studies, atopy is associated with increased DTH to atypical mycobacteria. Wong et al [2001] examined the relationship between tuberculin testing in 2600 children from the Hong Kong arm of the ISAAC study: asthma, hayfever and eczema risk was unrelated to tuberculin response, and the mean TT response was 3.4 mm in atopic and 3.3 mm in nonatopic subjects.

Two studies [Adams et al 1999; Ohrui et al 2000] have found sIgE levels to be elevated in TB patients compared to controls, and that this level falls with treatment, in parallel with an increase in tuberculin response.

Rarer diseases

The genetic disease Familial Mediterranean Fever (16p13) has been reported to be protective against asthma by some authors [Danon 1990; Ozyilkan 1994]. One study [Brenner-Ullman 1994] has suggested that heterozygotes also have a decreased incidence (OR=0.5, 95%CI=0.1, 2.1).

Conclusions about evolution

The evidence regarding allergy and cancer is confused, and a large predisposing or protective effect seems unlikely. In the case of parasitosis, a protective effect of a tendency to produce high levels of IgE seems more probable, though here too the evidence is patchy. Since allergic diseases were rare prior to the last 200 years (asthma perhaps excluded), an atopy gene was probably not deleterious until exposure to critical environmental exposures has become common more recently. These might be air pollution, dusty blankets, or removal of protective exposures such as early viral RTI or gut colonisation by lactobacilli.

Aetiology: conclusions

Overall, there is good evidence to implicate a number of the factors I have described as being causative of asthma. The key question is which factor is acting in the recent rapid increase in the incidence of asthma (and allergy). I think we can exclude air pollution as responsible (eg rural versus urban comparisons; East versus West Germany), as well as changes in breast feeding practices, maternal smoking [eg Ownby et al 1991] or diet. A few of the more likely possibilities are housing design (with an effect on infantile house dust mite exposure), smaller family size [Strachan 1989] leading to delayed exposure to viral pathogens (in which case infant day care should decrease asthma risk, but not necessarily wheezing), or perhaps a novel infectious agent (such as the Chlamydiae). Changes in sodium intake may be implicated in BHR, but cannot explain increased rates of allergic sensitisation. Many workers are concentrating on factors affecting the immune events occurring in the first three to six months of life, which leads onto the topics of the next chapter.