GENETICS OF ASTHMA AND ALLERGY: PEDIGREE ANALYSES

Early family studies of asthma and allergy

Wiener et al [1936] cites Sennertus (1650) who commented that his wife, three of her sibs and a niece all suffered from asthma. Salter [1860, p 109] states "Is Asthma Hereditary? - I think that there can no doubt that it is...the number of cases in which there is a family history of asthma is greater than will be found...on the mere doctrine of chance...Out of thirty-five cases...I find distinct traces of inheritance in fourteen...two cases out of every five". A first degree relative was affected in ten cases and he notes "...that several brothers and sisters in a family may be asthmatic without the parents having been so". In subsequent work [cited by Spain & Cooke 1924], he found a family history of asthma in 39% of a series of 217 asthmatics.

Cooke and coworkers [1916, 1924]

Robert Cooke performed two large studies of the inheritance of atopy, one reported in 1916 [Cooke & Vander Veer], the other in 1924 [Spain & Cooke]. The first study examined asthma, hayfever, urticaria, angioneurotic oedema and acute gastroenteritis in 504 subjects. In the second study, because some doubt had been expressed as whether the last three conditions were in fact part of the same atopic diathesis, only asthma and hayfever were used to define atopy in a further 462 individuals. All these subjects exhibited positive intradermal tests. A control series of 115 nonatopic probands was also recruited. Family history of atopy was determined by interviewing the proband "and in many cases, other members of the family".

A family history of atopy was present in 48.4% of cases in the first study, and in 58.4% of the 1924 study. Only 7% of the 115 normals reported a family history of atopy (crude OR=19; 95% CI 9-39). This was compared to the prevalence of asthma or hayfever - 3.5% - in a separate study where 119 medical students and nurses were surveyed on disease in themselves and their nuclear families. In the second study, the age of onset of the disease was found to be earlier in probands with both parents affected, compared to those where one or no parents were affected (see Table). The estimated recurrence risks, based on both studies, were 0.69 if both parents affected, 0.58 if one parent was affected, and from the normal series 0.02 if neither parent was affected.

Cooke felt the findings were most consistent with an autosomal dominant gene for atopy. This would give a penetrance of 0.6, and a gene frequency of 0.03. He did entertain the idea that it was due to a "...dominant and multiple [factors], the component parts of which are separately inherited", explaining the variability of age of onset and severity of disease.

Table 1. Age of onset and parental affection status for asthma and hayfever [from Spain & Cooke 1924].
No. parents affectedAge of onset
1-5 yo6-10 yo>10 yoTotal
Nil23 (12%)19 (10%)150 (78%)192 (41%)
One50 (21%)36 (15%)150 (36%)236 (51%)
Both18 (53%)9 (26%)7 (21%)34 (8%)

Schwartz [1952]

In 1952, Michael Schwartz published a description of his large family study using the Weinberg proband concordance method entitled "Heredity in bronchial asthma: a clinical and genetic study of 191 asthma probands and 50 probands with baker's asthma". This was performed during the period 1944-49 in Copenhagen. The 288 pages (of his doctoral thesis) include an extensive review of previous family studies of atopy and asthma.

The asthma probands consisted of those patients treated for asthma at the Medical outpatients clinic of University Hospital, Copenhagen, through 1943 and early 1944, and who lived in Copenhagen or environs. Of those eligible, 15% were either uncooperative or offered inadequate information about their families (illegitimacy etc). The probands with baker's asthma were also recruited from the same outpatients clinic, of which 10 cases were diagnosed in 1943-44. Schwartz gives the interesting statistic that 5 of the 271 asthmatics seen at this clinic in 1943 had died from asthma by 1946 (3 year mortality rate 1.8%, SE=0.8%).

The control series of 200 individuals were recruited from the outpatient clinic of the National Poliomyelitis Society (101 individuals), the surgical wards and outpatients of the municipal hospital (77 individuals), and was supplemented by 22 medical students and 2 masseuses. The controls were approximately age- sex matched for the cases. Leigh and Marley [1967, see below] have criticised the choice of controls, presumably as being unrepresentative of the general population. Schwartz comments that he had great difficulty in collecting an adequate number of controls, which explains the heterogeneity of sources. Since susceptibility to poliomyelitis is not known to be associated with atopy, as would be attendance at medical school, it seems unlikely that any great confounding had occurred. The use of hospital controls, coming from the same catchment, is also acceptable [Rothman 1986]. The author felt, but did not test, that cases and controls were matched on socioeconomic status.

The family material comprised questionnaire, and interview/clinical data from cooperative subjects. Eligible relatives included all first degree relatives, grandparents, and the sibs of the parents. The author personally interviewed and examined 2352 relatives, and obtained mailed questionnaire data from 1463 further subjects. Data were not obtained from 574 live individuals; this proportion did not differ between cases and controls (and by sex þ23=3.2, P=0.36). The ages of the various classes of relatives were comparable. Families were ascertained singly. There was an average of 16 relatives per proband of whom 8.7 were living; again no significant differences were found between cases and controls.

The asthmatics underwent examination of peripheral blood including eosinophil count, intracutaneous or scratch test for a number of allergens (usually 22) including dust collected from the patient's home. These along with the history were used to dichotomise the asthma cases into atopic and nonatopic. Atopic relatives and a small number of willing healthy relatives also underwent skin atopy testing and eosinophil count.

As would be expected, asthma, seasonal and perennial rhinitis, atopic dermatitis, and to a lesser extent urticaria were found to be associated in the same individuals. Asthma (Table 5.2), seasonal allergic rhinitis, "vasomotor" ie perennial rhinitis, and Besnier's prurigo (atopic dermatitis) were more common in toto in the relatives of asthmatics. It might be noted that the combined Mantel-Haenszel odds ratios for first and second degree relatives differ from each other by a factor of two - roughly consistent with an additive genetic model of transmission. I confirmed this using phi modelling (Table). The trait prevalence under the additive Phi model was 1.0%, which agrees well with Schwartz's estimate of 1.1%. His estimate for the incidence of atopic disease in the general population was 3.2% (SE=0.5%), significantly lower than that reported by modern studies. As noted earlier, the incidence of atopic diseases such as asthma have been reported to be lower in Scandinavian countries [Gregg 1986].

The author then compared the incidence rates for atopic disease in the asthmatic and control kindreds using person-years to reported age of onset for survival analysis. Both asthma and perennial rhinitis occurred significantly more often in the asthma families. The risk for eczema was found to be significant only for relatives of atopic asthmatics, while seasonal allergic rhinitis (hayfever, an uncommon disease in Denmark) did not differ for any group. He also reports a significant higher age-adjusted risk of atopy in siblings of asthmatic probands than in parents or offspring, and found that this could be explained either in terms of autosomal recessive or dominant inheritance, given the low incidence rates found. He comments that a higher population incidence with the given observed risks would support the recessive hypothesis. Examination of particular pedigrees however was supportive of a dominant gene with incomplete penetrance, the latter being estimated at 40% (60-96% in the siblings, and 30-45% for parent- offspring and second degree relatives). This difference in penetrances was thought suggestive of epistasis ("polymerism of the genotypic environment [of the monomeric]" major gene). The theory of Wiener, Zieve and Fries [1936] that early onset asthma occurs in homozygotes and later onset disease in heterozygotes was examined and found wanting [Schwartz, p 223 figure 8].

A critical feature of Schwartz' study is the finding that there was no difference in risk of atopic disease between the relatives of allergic and nonallergic asthmatics. Similarly, no differences were found for the prevalence of skin atopy or eosinophilia in these two groups in either unaffected relatives or those manifesting atopic disease. For asthmatic relatives of allergic and nonallergic asthmatic probands for example the prevalences of skin atopy were 20/30 and 29/39 respectively (P=0.5). It should be noted that cooperation with intracutaneous skin testing was not high in unaffected individuals from either asthma or control families.

The probands with baker's asthma and their families were not found to differ significantly from the other asthmatic series, thus confirming that a genetic predisposition was necessary to develop this occupational disease. Summarising, this study confirmed the existence of a genetic basis to atopy, and found that nonallergic and allergic asthma probably represented the same genetic entity.

Table 2. Occurrence of asthma in different classes of relatives of asthmatics and nonatopics observed by Schwartz [1952], and associated relative risks of affection.
Class of RelativeAsthma FamiliesControl FamiliesOdds Ratio (95% CI)
First degree relatives
Father6/1170/14016.3 (3.1-270)*
Brother18/2093/2176.7 (2.0-23.2)
Son7/991/755.6 (0.7-47)
Mother10/1300/16022.4 (5.8-385)*
Sister21/2183/1986.9 (2.0-23.6)
Daughter5/740/6310.0 (1.8-185)*
Combined odds (Mantel-Haenzel)8.0 (4.0-16.1)
Second degree relatives
Uncles14/3035/3743.6 (1.3-10.0)
Aunts23/3765/4345.6 (2.1-14.9)
Grandmothers1/591/801.4 (0.1-22.2)
Grandfathers3/490/24937.6 (5.5-740)*
Combined odds (Mantel-Haenszel)4.8 (2.5-9.3)
Total subjects108/1634 (6.6%)18/1790 (1.0%)

* Corrected for zero in one cell by Haldane adjustment, however lower limit produced for unadjusted data using Cornfield method.

Table 3. Genetic (binary correlation) model fitting to asthma in different classes of relatives of asthmatics and nonatopics observed by Schwartz [1952].
ModelLR X2d.f.PAIC
No familial aggregation (E)117.2190.00+79.2
Shared environment (CE)23.8180.16-5.1
Additive genes (AE)19.6180.35-16.4
ACE18.9170.33-15.1
ADE19.2170.32-14.8

Ratner & Silberman [1952] on allergy

These authors [1952] studied asthma, hayfever, eczema and recurrent urticaria in the first and second degree relatives of 250 allergic children, and also in a control group of 3000 medical students, nurses and doctors. Histories for a total of 6244 relatives of the probands were obtained via the proband's parents on repeated occasions as necessary.

A family history of atopy (the four conditions listed above) was found for 54.4% of the probands; a parental history in 22.4%. The prevalence of atopy in the first degree relatives of the probands was only 5.5% - though that in the parents was 11.8% and in the sibs 8.8%. Curiously, the authors state the prevalence of atopy in their control group was 10-13%. The authors made much of the fact that individual families with a high prevalence of atopy were relatively uncommon, and concluded by questioning any genetic mechanism as a cause for the familiality of atopy. The absence of comparable control family data weakens the claims made by the authors. Otherwise, they note that their results agree well with previous studies such as that of Spain and Cook [1924, see above].

VanArsdel & Motulsky [1959] on asthma and rhinitis

In this study [1959], 5818 freshmen attending the University of Washington in 1956-7 completed a questionnaire on personal and family history of allergic disease and underwent a routine screening examination at the University Health Centre. The bulk of subjects were 17-18 years of age, and consisted of 4110 males and 1708 females.

The lifetime prevalence of asthma or hayfever was 16%. The prevalence of asthma was 4.7%, and of hayfever 12%, the two coexisting in 2.7% of subjects. The overall male:female ratio for atopy was 0.8, but for asthma was 0.96, reflecting the suggested overreporting of allergic rhinitis by females. A family history of allergic disease was more commonly reported by the allergic subjects (see Table). Disease in second degree relatives was more likely to be reported by female subjects - there was no sex difference for reporting of disease in first degree relatives. The results for the different mating types are presented below (see Table). Although an item on membership of a twin pair was present in the questionnaire, no data were presented. Monogenic recessive models were fitted to the data, and were found compatible with a wide range of gene frequencies and penetrances. The authors concluded that the findings could as easily be due to polygenic inheritance.
Table 4. Reported prevalence of atopic disease in relatives of atopic and nonatopic college students [from VanArsdel & Motulsky 1959].
Class of RelativeAtopic subjectsNonatopic subjectsOdds Ratio (95% CI)
Parents383/1942*563/9694*3.9 (3.4-4.6)
First degree426/971779/48474.1 (3.5-4.7)
Number of probands9714847

*Number of parents at risk.

Table 5. Risk to proband of different mating types of reportedly atopic and nonatopic parents [from VanArsdel & Motulsky 1959].
Mating typeAtopicNonatopicNo. at risk
Atopic x Atopic29 (58%)2150 (1%)
Nonatopic x Atopic325 (38.4%)521846 (14.5%)
Nonatopic x Nonatopic617 (12.5%)43054922 84.5%)

This work is of course open to the criticism that information about the relatives' affection status is collected via the probands, a notoriously inaccurate method [Vogel & Motulsky 1986]. The prevalence in siblings and other specific classes of relative was not reported, limiting the analyses I can perform. Nevertheless, the risks agree broadly with those from other studies.

Conclusions

Of the studies carried out up to 1960, that of Schwartz [1952] is the best designed and executed, in that relatives were followed up very completely. The finding that occurrence of skin atopy in relatives of nonatopic asthmatic probands did not differ from that for atopic probands is a critical one. The estimates of recurrence risks from most of the studies seem to be consistent with one another, and a genetic cause for the familial aggregation seems most likely.

Later family studies of asthma and allergy

Schnyder [1960] on asthma

This large body of work [1960] encompassed a population survey, a community recruited twin series, and a family study of 361 families. Initially, the population of Zurich was surveyed. This gave prevalences for asthma 2-4%, atopic rhinitis 4-8%, and atopic dermatitis of 0.1-0.5%. Asthma, rhinitis and atopic dermatitis tended to coincide in the same individuals. In the family study, there was a similar aggregation of asthma, hayfever, and atopic dermatitis (see Table), but not contact dermatitis, urticaria or drug hypersensitivity. There was some evidence of specificity of type of atopy within families, with childhood atopic dermatitis being found more frequently in families of dermatitis probands than respiratory allergy probands (see Table 5.7). In the case of a family history of "pure" atopic dermatitis, ie without a coincidental history of respiratory atopy, the odds ratio was 14 (95% CI 5-39), but for a "mixed" family history, OR=7.9 (95% CI 3.9-16.0). The converse was also true in that a family history of asthma/hayfever was more common in respiratory atopy probands than those with uncomplicated atopic dermatitis (OR=3.7, 95% CI 1.5-9.0), but not if those kindreds where both asthma/hayfever and atopic dermatitis were reported were included (OR=1.3, 95% CI 0.8-2.0).

Skin atopy was tested in 191 probands by scratch followed by intracutaneous testing with five (mixed) aeroallergen preparations. There was no association between a family history of atopic disease and presence or absence of skin atopy in the proband: 77/131 skin test positive and 28/60 skin test negative, OR=1.6, 95% confidence interval= 0.9-3.0. The inheritance of atopy in the families "strongly supported the hypothesis of a single, autosomal dominant gene with reduced penetrance (40-50%)". "Sporadic" atopy was also thought to be largely genetically determined.

Table 6. Familial coaggregation of asthma, hayfever and eczema [from Schnyder 1960].
Family materialPrevalence in Affected linePrevalence in Unaffected lineOdds Ratio (95% CI)
Families with a unilateral history of asthma (42 families)31/162 rhinitics6/147 rhinitics5.6 (2.2-13.8)
Families with a unilateral history of rhinitis (63 fam.)23/214 asthmatics4/195 asthmatics5.7 (2.0-16.9)
Families with a unilateral history of asthma or rhinitis (102 fam.)9/391 with eczema1/348 with eczema8.2 (1.2-177)
Families with a unilateral history of atopic dermatitis (17 fam.)14/52 with asthma or rhinitis4/56 with asthma or rhinitis4.8 (1.4-16.7)

Table 7. Presence of different atopic disease in families of probands with asthma, hayfever and atopic dermatitis [from Schnyder 1960].
LineageProband affection status
Atopic dermatitis without asthma/hayfeverAtopic dermatitis with asthma/hayfeverTotal atopic dermatitisAsthma/hayfever
No family history* of atopic disease12 (46%)27 (30%)39 (33%)83 (34%)
Unilateral history* of:
Asthma/hayfever33235122
Atopic dermatitis5380
Both416209
Subtotal12 (46%)51 (56%)63 (54%)131 (54%)
Bilateral history* of:
Asthma/hayfever09927
Atopic dermatitis0000
Both2463
Subtotal2 (8%)13 (14%)15 (13%)30 (12%)
TOTAL26 (100%)91117244

* Disease present in sib, parent, uncle or aunt, grandparent, or in some cases great-uncle or great-aunt of parent(s) of proband.

McKee [1967] on allergy

Here [1967], results from 1000 patients at a general medical clinic are reported. Subjects were attending for a routine examination, and completed a questionnaire on asthma, allergic rhinitis, eczema and a number of other symptoms in themselves and their families. Only 34% of the subjects taking part in the study were definitely nonatopic, while 20% reported definite seasonal symptoms. The prevalence of asthma was 1.7%. As has been found repeatedly in the studies reviewed above, 50% of subjects with definite seasonal allergy gave a family history of atopy, as compared to only 18% of the normal subjects. Due to the selected nature of the subjects and the lack of detailed information about affection status in relatives, I will not discuss this further.

Leigh & Marley [1967] on asthma

These authors [1967] performed a community based family study using the Weinberg genealogical proband method. Subjects were recruited from two London practices (one being that of Dr John Fry) serving relatively socially homogenous suburbs (lower-middle and working class). All asthmatics from the smaller practice and a random sample of asthmatics from the other were chosen as case probands (N=55) while a random sample of age-sex matched (within 20 year age bands) non-asthmatics served as control probands (N=55). All available first and second degree relatives of the probands were then recruited into the study (250 1ø relatives of cases and 201 of controls; 289 2ø relatives of cases and 283 of controls). All probands and their immediate families were examined by the investigators, and all subjects within 25 miles of London were interviewed independently by a psychiatric social worker. Relatives outside this distance were sent a questionnaire. Eleven eligible probands (3 asthmatic and 8 control) refused to cooperate (90% cooperation rate) while 24 asthma and 25 control probands were rejected from the study due to incomplete information about relatives. More loss to followup of relatives of controls was found (9% lost or uncooperative for asthmatics and 16% for controls); this was partly due to the fact that more of the control families did not have a current address for 2ø relatives (33 v. 24).

Significantly more 1ø relatives of asthmatic probands were diagnosed as asthmatic than controls, but the risk to 2ø relatives did not differ in the two groups (Table). The authors gave a 31% risk of developing asthma by age 55 for 1ø relatives of asthmatics, and an 11% risk for 1ø relatives of controls. The fact that the ratio of prevalence in 1ø and 2ø relatives of asthmatics is greater than two suggests the presence of nonadditive genetic effects, possibly epistasis in the absence of age corrected data for parents and sibs.

Other more chronic respiratory conditions such as recurrent acute bronchitis or chronic bronchitis were also increased in the relatives of asthmatics, but not acute episodes of pneumonia or (acute) bronchitis. Asthmatic probands were more likely to suffer both allergic and vasomotor rhinitis, as were their relatives (see Table). The differences between 1ø and 2ø relatives were not significant (allowing these two to be combined with an overall OR cases v. controls of 1.8 (95% CI=1.2-2.5). However, this is suggestive of confounding of some type (age?), in that if there is a genetic component to hayfever, the risk should fall off with the distance of the relationship. Removing cases diagnosed as vasomotor rhinitis did not alter this picture.

Table 8. Occurrence of asthma in relatives of asthmatic and nonasthmatic probands [From Leigh & Marley 1967].
RelationshipCase groupControl GroupOdds Ratio
Parent13% (11/84)2% (2/79)5.8 (1.2-27.1)
Sib14% (18/132)0% (0/97)4-500*
Offspring12% (4/34)4% (1/25)3.2 (0.3-10)
1st deg Relative13% (33/250)2% (3/201)10.0 (3-33)
2nd deg Relative4% (11/289)4% (12/283)0.9 (0.3-2)

* Lower limit using Cornfield method, upper using Haldane approximation.

Table 9. Occurrence of rhinitis in asthmatic and nonasthmatic probands and their relatives [From Leigh & Marley, 1967].
RelationshipCase groupControl GroupOdds Ratio
Proband69% (38/55)9% (5/55)12.9 (4.5-37.4)
1st deg Relative20% (51/250)16% (33/201)1.3 (0.8-2.1)
2nd deg Relative18% (51/289)8% (23/283)2.4 (1.4-4.1)

Higgins and Keller [1975] on asthma and chronic bronchitis

This paper arises from the Tecumseh (Michigan) Community Health Study. Results from examination for respiratory disease in the families of all 9226 subjects (82% of the total community population) were reported. Disease was diagnosed on questionnaire and clinical criteria - chronic bronchitis using the MRC definition, asthma as noted in section 1.1, and allergic rhinitis as: "reported hayfever, sinus trouble or persistent runny or stuffy nose associated with:

  1. itching of eyes, nose or throat or burning, watering or redness of eyes;
  2. exposure to allergens; and
  3. a physician diagnosis of hayfever or allergic rhinitis."
Spirometry was recorded for all subjects. Familial aggregation of diagnoses and respiratory function were examined. Approximately half of the sample had both a sib and a parent participating in the study, and 89% one or more first degree relatives.

Significant parent-offspring correlations were noted for asthma, chronic bronchitis and allergic rhinitis (see Table). Similar aggregation of diseases was noted within sibships. There was no significant marital correlation observed. Males exhibited a higher prevalence of chronic bronchitis, and asthma for ages under 40 years. Age-sex-height standardised ventilatory function was also correlated within families. Significant parent-offspring and sibling correlations were observed, as well as a smaller marital correlation (approximately the same size as the marital correlation for stature, 0.16). Statistical adjustment for covariates such as disease and smoking status were not performed.

The relative risks for parent-offspring respiratory disease observed agree well with those from other studies. The association seemed to be strongest for children below the age of 40 years - an early onset for chronic bronchitis due to cigarette smoking - however the exact numbers in each cell of the table were not given. Numbers of children over 40 years of age were low, so possibly the pattern of associations might have extended to the older group, but was not detected due to inadequate power.

Table 10. Parent-offspring associations for chronic bronchitis, asthma and allergic rhinitis [from Higgins & Keller 1975].
ConditionMatingMale offspring (% of)Female offspring (% of)
>16 y.o.16-39 y.o.40+ y.o.>16 y.o.16-39 y.o.40+ y.o.
Chronic BronchitisNonenil6.313.8nil2.06.9
1 or 2nil17.316.7nil6.68.3
RR (p-o)--2.7*1.2--3.3*1.2
AsthmaNone7.48.33.24.14.915.4
1 or 218.312.2nil11.713.010.0
RR (p-o)2.5*1.5--2.9*2.7*0.65
HayfeverNone3.412.417.42.914.929.6
1 or 213.532.130.810.925.229.4
RR (p-o)4.0*2.6*1.83.8*1.7*1.0
Number of subjects7796601051779621110

* Significant at 5% level.

Fergusson et al [1983] on asthma and eczema

These authors [1983] describe the findings of the (New Zealand) Christchurch Child Development Study on transmission of atopy in 1110 children and their parents. The families were ascertained via a child born into the study cohort, and followed up on six occasions until the child was four years of age. Health data were obtained by structured interview, medical records and a health diary. Log-linear models which included parental (mother and father not being separately identified in the model) and proband asthma and eczema were fitted. The table below presents the crude relative risks associated with various family types.

Asthma and eczema were unsurprisingly associated in the child, as they were in the parents. For female children, although there was an association between parental eczema and eczema in the proband, there was no such association for asthma. For the males, where asthma tended to occur at an earlier age, parental asthma was associated with asthma in the child. The separate first order interactions for parental-offspring asthma and parental- offspring eczema were thought to suggest specific transmission of each disease (in addition one would presume to a general tendency to atopy).

Because families with one or two affected parents were not differentiated, as were those families where one parent reported eczema and the other asthma from those where the parent reported both, genetic hypotheses cannot be further examined, and conclusions about the patterns of familiality of these diseases should be made with caution.

Dold [1992] on atopic disease in German children

Dold et al [1992] reported a similar study of 6665 schoolchildren (age 9-11 years) in Munich and Bavaria. The questionnaire used (58 items) was based on the ATS-DLD. The end points analysed were physician diagnosed asthma (or recurrent wheezy bronchitis), allergic rhinitis/hayfever, and atopic dermatitis - either a physician diagnosis of eczema, neurodermatitis, or "itching skin lesions in typical locations". The cumulative incidence of asthma in probands with a least one sibling was 7.6%, of allergic rhinitis 9.5%, and atopic dermatitis 19.5%. Rates were higher where the parents smoked, or were better educated. The parent-offspring concordances were almost identical to those obtained by Fergusson et al [1983], and I have tabulated comparable results below (Table), which confirm transmission of specific atopic conditions.

Table 11. Four-year cumulative incidences and odds ratios for parent-offspring and gender associations for asthma and eczema [adapted from Fergusson et al 1983].
ConditionAffection statusProband AsthmaProband Eczema
Parental Asthma1 or 2 (N=202)11.4%30.7%
None (N=908)5.9%20.6%
OR (p-o)2.0 (1.2-3.4)1.7 (1.2-2.4)
Parental Eczema1 or 2 (N=249)8.4%32.5%
None (N=861)6.5%19.5%
OR (p-o)1.3 (0.8-2.2)1.7 (1.3-2.4)
Sex of ProbandMale (N=560)9.8%22.3%
Female (N=550)4.0%22.5%
OR (Male sex)2.6 (1.6-4.4)1.0 (0.8-1.3)
No. of affected probands77 (6.9%)249 (22.4%)

Table 12. Cumulative incidences and odds ratios for parent- offspring and gender associations for asthma and eczema [adapted from Dold et al 1992].
ConditionParents AffectedProband AsthmaProband Eczema
Parental AsthmaOne (N=166)15.1%20.6%
None (N=3994)6.4%15.0%
OR (p-o)2.6 (1.7-4.0)1.5 (1.0-2.2)
Parental EczemaOne (N=302)6.3%37.7%
OR (p-o)1.0 (0.6-1.6)3.4 (2.6-4.4)
Sex of Proband (in Atopy families)Male (N=1914)7.8%14.0%
Female (N=2080)5.2%15.9%
OR (Male sex)1.5 (1.2-2.0)0.9 (0.7-1.0)
No. of affected probands77 (6.9%)249 (22.4%)

Åberg [1993]

This massive questionnaire survey [1993] obtained information on 19814 Swedish children and their parents (a 92% response rate). As in other such studies, the mother was the informant in almost all cases. This may explain a two-fold excess of atopy in mothers compared to fathers - urticaria, eczema, hayfever and asthma ("one or more episodes of heavy breathing and/or wheezing...not...explained by other disease"), as well as a marital correlation for atopy. The odds ratios for atopy in the mother-child and father-child did not differ significantly (Mother OR=2.8, 95%CI=2.6-3.0; Father OR=2.75, 2.5-3.0). There was evidence similar to that above of trait-specificity in parent-offspring association - the same-trait odds ratio being twofold higher than that for the pooled odds ratio for other atopic traits. For example, for asthma in the parent and offspring the odds ratio was 4.8, and for other atopy in the parent and asthma in the child, 1.8.

Duffy and Mitchell [1993] on asthma and cough

In this paper [1993], I analysed data collected in 1980 by Dr Mitchell on 4549 Queensland children (aged 7 years and 11 years). Again, the mother (usually) completed the questionnaire, on this occasion the TAFQ. Mothers reported a history of wheeze for themselves more frequently than for the father. Frequent wheezing (every three months or more often) was predicted by a history of maternal wheeze - OR=4.1 (95%CI=2.9-5.9), and to a lesser extent by paternal wheeze (OR=2.4, 95%CI=1.6-3.4), while hayfever was a less stronger predictor (OR=1.7, 95%CI=1.2-2.4). We could not demonstrate a gradient of severity within probands with an increasing number of parents affected (Table), which I interpret as evidence against a multifactorial threshold model. Such a gradient was demonstrable for the risk of frequent cough (in children with no history of wheeze), where we also noted that a parental history of "frequent bronchitis" was a stronger predictor than parental wheeze. It is interesting to compare the first line of results in Table with those found by Åberg [1993] for atopy (nil atopic parents, 18% of offspring atopic; one, 38%; two, 52%).

Table 13. Frequency of episodes of wheeze in children tabulated by number of natural parents with past history of asthma or wheezing (Trend X2(1 df)=1.15, P=0.28). From Duffy and Mitchell [1993].
Number of Affected ParentsNoneOneBoth
Ever Wheeze396 [15.4]*358 [35.9]81 [47.1]
One or more attack per month46 (11.6)**54 (13.2)10 (10.9)
One attack per 3-12 months132 (33.3)121 (33.8)29 (35.8)
Less than one attack per year218 (55.0)183 (51.1)42 (51.8)

* Square brackets denote percentage of all children from mating type (Unaffected x Unaffected, Affected x Unaffected, or Affected x Affected).
** Round brackets percentage of affected offspring of mating type.

Sibbald and coworkers [1980] on asthma and bronchitis

Sibbald, Horn, Brain and Gregg [1980] reported results of a Weinberg proband concordance study on asthma and skin atopy that examined the nuclear families of 77 asthmatic and 87 nonasthmatic children. The children were selected from the patients of a single general practice in south-west London. Children with a history of solely RTI-associated wheeze ("wheezy bronchitis") were excluded from the study. Skin atopy to a number of common allergens was measured in all probands and in all accessible relatives. Atopy was defined as one or more reactions 2 mm (2 cm in the text) in diameter ("in the absence of any equivalent reaction in the control solution"). Asthma in the probands was diagnosed by clinical examination, while the presence of asthma, hayfever and eczema in relatives was diagnosed by interview and examination of medical records.

The mean age of the probands was 6.3 years (asthmatics 7.1 y, controls 5.6 y). Asthmatic probands had an increased risk of hayfever, eczema, skin atopy, and family history of asthma than did the controls (see Table). The prevalence of asthma in the relatives of atopic asthmatics was slightly higher than that among relatives of nonatopic asthmatics (see Table), but this difference was not significant. The authors also noted that "in relatives of asthmatics, the prevalence of atopic asthma exceeded the prevalence of nonatopic asthma, irrespective of the atopic status of the proband", but this difference also was not significant at the 5% level (X2(2 df)=4.88, P=0.09 compared to distribution in control families).

There was a marginally significant difference in prevalence of skin atopy as defined in this study in relatives of atopic asthmatics compared to all controls (see Table). Closer examination reveals that this is due largely to the relatives of the nonatopic controls, where the prevalence is significantly lower than that for the relatives of the atopic asthmatics. The difference between risk to relatives of atopic and nonatopic asthmatics is almost significant (z=1.75, P=0.07), but the small number of nonatopic asthmatics in the study prevents more being made of this.

In the conclusion, the authors comment that the results agree well with those of Marley & Lewis [1967]. They interpreted the fact that recurrence risk of atopy was not associated with asthma status of the proband as suggesting that asthma and atopy are inherited independently, but that atopy increases the risk of expressing the asthmatic trait. Since the prevalence of skin atopy was high in the controls (44%), compared to that obtained in population surveys, they warned against overinterpretation of the data with respect to atopy.

The increased risk of asthma in parents compared to that in siblings might be due to the young age of the probands and thus presumably siblings, combined with a continuing incidence rate at later ages. Another explanation might be a cohort effect, but this would be in the opposite direction to the secular increase in asthma incidence usually reported. A further point to note is the exclusion of "wheezy bronchitis" from the study group, which presumably was also extended to the relatives, as no data for this entity were presented. Inclusion of RTI associated wheeze in the analysis might have shed a different light on the difference in asthma rates in the parents and children, as well possibly on the nonatopic versus atopic dichotomy.

Table 14. Prevalence (fraction) of asthma in relatives of probands and recurrence risks (odds ratio) of atopic (OR1) and nonatopic asthma (OR2) compared to all controls [Adapted from Sibbald et al 1980].
Type of RelativeProband Affection StatusRecurrence Risk
Atopic AsthmaNonatopic AsthmaAtopic ControlNonatopic ControlOR1 (95%CI)OR2 (95%CI)
Parents18% (23/128)15% (4/26)3% (2/76)5% (5/98)5.2 (2.2-12.6)4.3 (1.2-16.0)
Siblings9% (11/124)4% (1/26)4% (2/53)3% (2/75)3.0 (0.9-9.8)1.2 (0.1-11.6)
1st deg relatives13% (34/352)10% (5/52)3% (4/129)4% (7/173)2.8 (1.4-5.7)2.8 (0.9-8.5)

Table 15. Prevalence of types of asthma in first degree relatives of atopic and nonatopic asthmatic probands [from Sibbald et al 1980].
Type of Asthma in 1st degree relativesProband affection statusAssociation
Atopic AsthmaNonatopic AsthmaAtopic ControlNonatopic ControlOR1 (95% CI)OR2 (95% CI)
Atopic10% (10)*14% (2)3% (1)3% (2)3.7 (1.0-13.9)5.6 (0.8-37.0)
Nonatopic3% (3)0% (0)3% (1)3% (2)1.0 (0.2-5.2)1.2** (0.1-26.3)
Total101143569

* Percent (number).

** Using 0.5 to replace 0 in one cell.

Table 16. Prevalence of skin atopy in first degree relatives of asthmatics and controls [from Sibbald et al 1980].
Type of relativeProband affection statusAssociation
Atopic AsthmaNonatopic AsthmaAtopic ControlNonatopic ControlOR1 (95% CI)OR2 (95% CI)
Parents 61% (34/56)38% (3/8)55% (11/20)45% (18/40)1.6 (0.8-3.4)0.6 (0.1-2.9)
Siblings60% (27/45)33% (2/6)60% (9/15)38% (11/29)1.8 (0.8-4.2)0.6 (0.1-3.7)
1st degree relatives60% (61/101)36% (5/14)57% (20/35)42% (29/69)1.7 (1.0-3.0)0.6 (0.2-2.0)

Kuehr et al [1993] on specific allergy

This interesting study looked at parent-offspring concordances for specific allergen sensitisation in 302 German nuclear families [1993]. Asthma or hayfever was present in 16% of the children (age 7-16 years of age), and approximately the same proportion of the parents (median age 40 years). The prevalence of skin atopy (a positive response to SPT for D pter, cat, mixed grass, or birch pollen) was 39% of children and 27% of parents - in keeping with their respective ages.

In logistic regressions, maternal atopy was a stronger risk factor for sensitisation in the child than paternal atopy. In the cases of grass, birch and cat, maternal sensitisation to that allergen was significantly associated with sensitisation in the child to that allergen (OR=2.5-4.1; for D pter, OR=1.8), when adjusted for positivity to any of the remaining three allergens. Positivity to any of the other three allergens was a significant associate for grass, birch and D pter (OR=2.7-3.7), but not cat (OR=1.6). I would prefer to see analyses using all four allergens in the mother simultaneously - presumably the numbers were too small. In addition, the marital correlation for atopy was not described - this might fuel shared environmental hypotheses for these findings. The alternative hypothesis is of a Carter effect, or a maternal effect - such as transplacental sensitisation of the fetus.

Smith & Cadoret [1990] on asthma and rhinitis

This is the only adoption study of atopic disease that I am aware of [1990] (although data has been collected on allergic disease as part of the Colorado Adoption Project). The prevalence of asthma and allergic rhinitis in subjects adopted into an allergic family (114 out of 230 studied families) was compared to that in probands from 73 "allergic" families whose mode of ascertainment was not described. Diagnosis was made by questionnaire (based on the WHO questionnaire, which closely follows the ATS-DLD-78).

Asthma was more common in adoptees than natural children (OR= 1.5, 95% confidence interval 0.7-3.3), while atopy was more common in the natural offspring of allergic families (OR=1.4, 0.8- 2.6). The sample size means the power to detect a 2.5 fold increase in risk for atopy would be 80%, so either the recurrence risks reported in some of the studies examined above are largely due to shared exposure to environmental factors as well as shared genes, or some form of confounding is present. Since many of this study's details are unknown, this point cannot be further addressed.

Mrazek, Pauls et al [1989] on asthma

Segregation analyses using POINTER of 145 families ascertained via a female asthmatic proband were reported in abstract form by these authors [1989]. The environmental model was rejected. Under the polygenic model the heritability was 0.96 (ASE=0.05). Neither the solely polygenic and monogenic models fitted significantly better than the mixed model, but the likelihood for the monogenic model was higher, so this was chosen as the preferred model. A recessive inheritance was found to best fit the data, with a homozygote penetrance of 62%, and 1% for the other genotypes, a gene frequency of 0.22, and a proportion of phenocopies estimated at 17%. This gave a population prevalence for asthma of 3.5%.

Jenkins et al [1997] on asthma

This is a segregation analysis of data from the Tasmanian Asthma Survey (1968). The index persons were all children born in 1961, and questionnaires were completed on their behalf by a parent in 7394 families (out of 8683). The definition of asthma was a positive response to the question "[Has he/she] Have you ever suffered from attacks of asthma or wheezy breathing?".

Asthma was reported in 16% of index cases, 11% of siblings and 11% of parents. The odds of asthma in the proband increased 3.4 fold if only the mother was affected, 2.8 fold if only the father, and 7.2 fold if both were affected.

The best fitting model was a class A regressive logistic model modified to include a sibship effect. Among the Class D models, the general transmission model (ie non-Mendelian) gave the best fit.

Manjavidze et al [1990] on allergy

This Russian study [1990] entailed study of the families of 225 allergic probands (age 6 months to 6 years) and of 1545 unselected age matched controls. As usual, the fully penetrant Mendelian hypotheses were rejected. The authors concluded that either a multifactorial or a single low penetrance dominant gene could explain the segregation patterns observed.

The European Community Respiratory Health Survey Group [1997] on asthma

This is a massive collaborative epidemiological study of asthma involving 56 centres throughout Europe and the world. Individuals from 13963 families in 15 countries completed a standardised interviews (1990-1992) on asthma status in themselves and their immediate family (total of 75392 individuals reported on). Asthma was apparently defined as an affirmative answer to the question "Have you ever had asthma?"

The overall lifetime prevalence of asthma in the index subjects was 6.9% (ages 20-48 years), and ranged from 2.1% to 16.2% across the different centres. The Australian centre had the second highest prevalence of 14.7% and the highest prevalence in the parents of the index (10.6%). Overall, the odds of asthma increased 2.9 fold if the father was reported affected, and 3.2 fold if the mother was affected, but there was between-centre heterogeneity of risk. No information on rates or risk to siblings was presented.

A Class A regressive logistic model was fitted to the data. The Mendelian codominant model (frequency of increasing allele A, 0.24; AB penetrances 4-13%; AA penetrances 32-70%) fitted as well as the general model.

Conclusions

The most striking feature that I note is the consistency of the findings in several different countries and populations. The most attractive example of this is found by comparing Fergusson et al [1983] with Dold et al [1992]. Another is the ratios comparing recurrence odds in first degree relatives of asthmatics and nonasthmatics (Figure 5.1).

Lung volumes

There are several published twin and family studies of FEV1 and FVC, but the recent paper by Holberg et al [1998] is interesting in that it compares families stratified on presence/absence of asthma. The 309 pedigrees (1163 individuals) were taken from the Tucson Children's Respiratory study. Much is made of a difference in the father-offspring and mother-offspring correlations across the two groups.

Table 17. Familial correlations for standardised FEV1 versus family history of asthma. From Holberg et al [1998].
TypeFH PositiveFH Negative
rNrN
Father-Off0.0631540.231391
Mother-Off0.2581540.177391
Sib-Sib0.398910.263252

Bronchial hyperresponsiveness

Townley, Elston and others [1986] on bronchial responsiveness

Townley and coworkers [1986] reported the results of a complex segregation analysis, using the method of Elston and Stewart, designed to detect a major gene determining bronchial responsiveness. The study was undertaken specifically to examine previously mentioned bimodal distribution of bronchial responsiveness in relatives of asthmatics, a finding consonant with the effects of a dominant major gene. The subjects used were 83 families from the Natural History of Asthma study. The 51 group I families (467 individuals) were ascertained via an asthmatic proband (age 6 to 25 years), while the 32 Group II families (N=291) were ascertained through nonatopic individuals. Most (26 families) of the Group II families were selected by the additional criterion that there was no history of atopic disease in three generations. Six families were ascertained through probands carrying the PiZZ alpha-1-antitrypsin genotype. In the latter case, the authors have not stated if these were asthmatics or not. This is germane as there is a reported excess of asthma in the relatives of homozygotes.

Methacholine provocative testing was performed using the two protocols used by this group. The resulting methacholine area-under-curve for 35% drop in FEV1 (Area35) were regressed on age, sex and recent respiratory tract infection. Other measures such as skin atopy were not included in the analyses. Correction for ascertainment was made using an exponential model: Both Area35 and age standardised (regressed on age and age2) Area35 were examined in the segregation analysis using a sex- limitation model.

Methacholine area exhibited a bimodal distribution for the Group I families, as did the Group II families to a lesser extent. The Mendelian and environmental models were fitted separately to each group of families. In both groups the Mendelian hypothesis was rejected, as was the (random) environmental hypothesis (see Table).

Table 18. Likelihood ratio X2 tests for Mendelian and environmental hypotheses for segregation of methacholine response tested under Elston & Stewart model [from Townley et al 1986].
HypothesisUnadjusted for AgeAdjusted for Age
Group I Group IIGroup IGroup II
Genetic: tau(AA)=1; tau(AB)=0.5; tau(BB)=0X2256.7538.51203.1729.10
df32,3*33
Environmental: tau(AA)=tau(AB)=tau(BB)X26.3317.756.2511.16
df21,2*22

* On bound.

** tau(AA) is the transmission probability for A from the parental genotype AA.

A further analysis of the Group II families excluding those with a history of RTI in the previous month was again marginally more supportive of the environmental hypothesis (X22=4.86; P=0.09) than the Mendelian hypothesis (X21,2=5.93, P=0.05), but the sample size is smaller (exact size not stated).

The authors concluded that there was evidence for familial aggregation of methacholine response, but that this was not due to a single gene, and may have been due to shared family environment. The latter possibility in the Group II families may have been due to familial transmission of viral RTI, as suggested by the results obtained after excluding those reporting such an event. There was a further fault in the analysis, in that the genotypic means for Group I and Group II families were different, suggesting that inadequate ascertainment correction, or the strict selection of Group II families, had biased the results.

A number of other criticisms could be made of this study. Firstly, it would seem sensible to have included a measure of atopy in the analysis - skin prick test data has been presented by Townley's group in most of their other publications from this ongoing program of research, including their twin study of asthma [Hopp et al 1984; see below]. Atopy is strongly associated with bronchial hyperresponsiveness, and of course is genetically determined. Similarly, other forms of genetic analysis may have been profitably applied to the data, such as path analysis or the mixed model. In view of the heritability estimates these authors had obtained for bronchial responsiveness in their earlier twin study, some extension would have been doubly warranted.

Longo et al [1987]

This is one of a number of small studies that have confirmed the existence of the earlier mentioned bimodal distribution of bronchial responsiveness in the relatives of asthmatics. Longo et al [1987] examined bronchial responsiveness to carbachol, skin atopy and total sIgE level in 50 asthmatics, the nonasthmatic parents of 40 asthmatics, and 70 nonasthmatic, family history negative controls. No subject had suffered a RTI in the previous six weeks. Bronchial responsiveness was measured as a percentage: the numerator being area-under-curve of the dose response curve (percentage fall FEF25-75 versus volume of breath units of carbachol 0.1 mg per 10 l inhaled); denominator a 50% fall FEF25-75 multiplied by 80 vital capacity breath units. Skin atopy was defined as a 3 mm or greater response to any of six common allergens.

Bronchial responsiveness exhibited a bimodal frequency distribution in the control group, and even more so in the parents of asthmatics. A unimodal skewed pattern was found for the asthmatics. Bronchial hyperresponsiveness was diagnosed in 10% of the controls, 50% of the parents, and 100% of the asthmatics, using the authors' usual cutting value of 60%. There were no sex differences in the parents. BHR was absent in only 6 of the 40 sets of parents of asthmatics. Skin atopy (OR=1.1, 95%CI=0.5- 2.5) and sIgE level (OR=0.7, 95%CI=0.3-1.7) did not discriminate between the hyperresponsive and normal parents or controls (see Table).

Table 19. Bronchial responsiveness and atopy [from Longo et al 1987].
Test performedHyperresponsiveness
presentabsent
Prick testpos1021
neg3782
sIgE>100IU/ml823
<100IU/ml3980

Hopp et al [1988]

This study [Hopp et al 1988] examined similar data to that just described, measuring bronchial responsiveness, skin atopy and sIgE level in the nonasthmatic parents of 31 asthmatics. Again, a bimodal frequency distribution was exhibited for bronchial responsiveness, allowing the sample to be divided into a low-minimal responsiveness and a marked- moderate responsiveness group. These groups did not differ in terms of age, sex ratio, aggregate skin atopy score, sIgE level, recent RTI or smoking status.

Bruderman et al [1987]

Bruderman, Cohen, Shachor and Horowitz [1987] reported similar findings in a study of 28 parents of asthmatic children, and a smaller number of controls. Allergic rhinitis was present in 9 of the 28, and 3 of these reported dyspnoea accompanying episodes of acute rhinitis. All underwent a methacholine provocation test.

Decreased FEF50 or FEF75 was present in a number of the parents. The subjects were divided into an group with diminished resting lung function or rhinitis (N=17), and a normal group. These groups differed significantly on methacholine provocation, with the first group having a PD25% SGaw of 0.17 mg Mch and the second group a PD25 of 1.3 mg Mch. The three rhinitics who gave a history of episodic dyspnoea subsequently developed full blown asthma; they exhibited greater bronchial responsiveness than the other subjects. Bruderman, Cohen and Schachter [1990] have also presented the results of a study of 24 nonasthmatic children of asthmatic parents (and 7 control children). Again, methacholine provocation was performed. All but three of the children of asthmatic parents demonstrated bronchial hyperresponsiveness with a mean PD35%þSGaw of 2.2 mg Mch. A subgroup of 13 hyperresponsive children (8 with skin atopy, 5 without) were then treated for 6 weeks with cromoglycate via metered aerosol. Bronchial responsiveness decreased significantly in the atopic children treated with cromoglycate, but not in the nonatopic children.

The authors combine the results of these two studies to argue that asymptomatic bronchial hyperresponsiveness occurs in the relatives of asthmatics, is associated with atopy but can occur in its absence, improves with asthma therapy, and might develop into late onset asthma with time. These results are very suggestive, but again numbers are small.

Clifford et al [1987]

In this study [1987], bronchial responsiveness (protocol of Yan et al) was measured in 29 offspring of an asthmatic parent (the other being nonasthmatic) and 30 parents, skin atopy (3 allergens) in 32 pairs, and questionnaire data was collected from these and a further 18 parent-offspring pairs. These results were compared to data from the authors' population based bronchial responsiveness survey for children of the same age (7 years). Parent-offspring pairs had been recruited for a parental history of asthma through a previous study carried out in an antenatal clinic (50% response rate to the questionnaire, and 30% to clinical phase).

The mean age of the children was 6.4 years. There were 10 children who had previously been diagnosed as having asthma. Nocturnal and diurnal dyspnoea, cough and morning tightness was more common in the children (66%) than the population controls, but not wheeze (!). Bronchial hyperresponsiveness (defined as a PD20 <6.4 umol Mch) was present in 93% (26/28) of the parents and 45% (13/29) of the children. The population prevalence of bronchial hyperresponsiveness was 29.4% (not significantly different from the offspring results). While 88% (28/32) of the parents exhibited one or more positive skin prick test results (wheal diameter>3 mm), this was present in only 19% (6/32) of the children. In the control group the prevalence of skin atopy was 24.5%. Bronchial responsiveness was significantly associated with skin atopy in the children (5/6). Of the 8 nonatopic offspring with bronchial hyperresponsiveness, 6 reported respiratory symptoms. Concluding, there are weaknesses associated with this study, especially with respect to self selection for significant symptoms as a bias.

Verity et al [1984]

Verity, Vanheule, Carswell & Hughes [1984] undertook a sib-pair study of bronchial responsiveness and skin atopy using 75 index asthmatic children and 75 siblings, a number of whom also suffered from asthma. All underwent an exercise provocation test (6 minutes on a treadmill running at a pace associated with a heart rate of 160 bpm), and skin prick testing with 5 common allergens.

The median age of the subjects was 9 years. A history of asthma was present in 15% of the siblings and wheezy bronchitis in a further 15% (total 15/37 boys and 8/38 girls). A parental history of atopy was elicited in 44% of the pairs, and a history of an affected sibling of the index case in 53%.

If a 10% drop in peak flow was taken as a positive exercise provocation test, 83% of the index cases were positive, as were 54% of the male siblings and 26% of the female siblings (P<0.05). Only half the positive siblings gave a history of wheeze, however. Skin atopy (one or more wheals greater than 3 mm diameter) was not significantly associated with bronchial hyperresponsiveness. Skin atopy was present in 15/30 siblings with positive exercise test, and 28/45 sibs with normal reactivity (P=0.3). No sex differences were noted for either sibling skin atopy or bronchial hyperresponsiveness, though the male index cases exhibited larger mean wheal diameter for house dust mite allergen than did the females. In conclusion, a male preponderance for bronchial hyperresponsiveness was not associated with an excess of skin atopy; skin atopy overall was not associated with hyperresponsiveness.

Kok et al [1998]

Several papers that describe segregation analyses of BHR are found below, as they were performed along with analyses of sIgE. The paper by Kok et al [1998] merely looks at recurrence risks for BHR (methacholine) in relatives of 115 children with allergic rhinitis. One third of the probands exhibited BHR, and BHR in the parents and siblings was more likely if the proband was BHR+ (OR=2.6, naive 95% CI=1.4-4.8). Interestingly, atopy was not increased among BHR+ relatives compared to BHR- relatives.

Conclusions

Firstly, the bimodal distribution of bronchial responsiveness is a real entity. However, this is probably not independent of skin atopy, and how this occurs in the relatives of asthmatics. This is because most large community based studies have shown that skin atopy is positively correlated with bronchial responsiveness (albeit not strongly), and because asthma and atopy coaggregate in families. In view of either small sample size, inability to allow for age confounding, or failure to include atopy as a covariate [Townley et al 1986], it cannot be stated with any certainty that family factors other than "atopy genes" are operating.

Serum immunoglobulin E

Marsh, Bias & Ishizaka [1974] on IgE and SPT

This in many ways pivotal paper [1974] described analyses of total sIgE level and skin atopy in samples of unrelated atopic subjects and 28 nuclear families (56 parents and 108 children) ascertained via an index child in each family allergic to grass and/or ragweed pollen. Blood samples were collected out of pollen season in the majority of cases and quantitative intradermal skin testing performed. The distributions of log transformed total sIgE level (logIgE) in the 205 unrelated atopics were compared to those found in a control group of 106 nonatopics, and a discriminating threshold of 95 U/ml chosen, in good agreement with previous reports [Gleich et al 1971]. This was used to dichotomise the family members into low and high IgE producers.

Examination of the various mating types was found to be consistent with an autosomal recessive pattern for high IgE, confirming the findings in the authors' pilot family study (see Table 5.19). The gene frequency for the high IgE allele was estimated from population data at 0.52, and 0.55. The frequencies of the mating types were consistent with this, given the method of ascertainment (X22=1.21). An approximate expectation for number of completely unaffected families (mating types rr x rr and rr x rR and 50% of rR x rR - taking mean number of offspring as two) in the general population under this model is 40%.

Table 20. Mating types and affection status of offspring (high or low sIge) arranged for recessive hypothesis [from Marsh et al 1974].
Mating typeNo. of familiesNo. of childrenNo. affectedExpected No.*
At least one offspring with high IgE
Low x Low10411815.3
Low x High9312117.5
High x High5252525.0
No offspring with high IgE
Low x Low120
Low x High390

* Under fully penetrant autosomal recessive hypothesis.

The concordance of skin reactivity to specific allergens and HLA haplotype was then examined in sibling pairs. No significant differences in concordance at the differing number of haplotypes shared were found. Concordance was increased by contrast when the IgE levels in each sib were both high. This suggested that any association between HLA type and reactivity to specific allergens was far less important than the gene controlling overall IgE responsiveness (but see below).

Turner et al [1974] on IgE in Busselton

As part of the Busselton Children's survey, data on the occurrence of atopic disease (N=1598) and sIgE level (N=1061) were collected [1974]. Parental disease status and sIgE was available for 555 of these. The diagnoses of asthma, hayfever, wheezy bronchitis and eczema were made by single questionnaire items (eg "Has the child ever had asthma?").

A history of wheeze was present in 20% of children and 28% (20%) of adult men (women). Hayfever was reported for 10% of children and 20% of adults. sIgE was elevated in both children and adults suffering wheezing or hayfever. The odds ratio for parent-offspring transmission of wheeze was 1.7 (95%CI=1.2-2.2) and for hayfever, OR=2.4 (95%CI=1.8-3.4). There were no effects of parental sex (LR X2(2 df)=2.1 for wheeze, 0.6 for hayfever). There were correlations between parents and offspring sIgE levels (unfortunately only Spearman rank correlations were presented - Table). The authors interpreted the fact that cross-sex parent-offspring correlations for sIgE were higher than same-sex correlations as suggestive of X-linked inheritance. A similar pattern of correlations was found by Lebowitz et al [1974], and the spousal correlation by Meyers et al [1986] (see below).
Table 21. Familial correlations (Spearman rank) for IgE. All correlations above 0.1 are significant at alpha=0.05, save that for sister pairs [from Turner et al 1974].
FatherMotherSonDaughter
Father IgE1.00
Mother IgE0.111.00
Son IgE0.040.120.19*
Daughter IgE0.160.080.220.14*

* Correlation between siblings.

Lebowitz, Barbee & Burrows [1984] on IgE and SPT

This study [1984] was part of the Tucson Epidemiology Study of Airway Obstructive Disease. A representative stratified sample of 344 nuclear families (N=1359 individuals) completed the ATS-DLD-77. Total sIgE level was determined for 90% of all subjects, and skin prick testing (to 5 locally important antigens and antigen mixtures) performed on 95%. A skin test index weighting wheal size and allergen importance was calculated. MANOVA was the main method used for the genetic analyses.

Parental history of medically diagnosed asthma was associated with an increased risk of asthma in the offspring, with a relative risk of 3.5 (95% confidence interval 2.0-6.2). A weaker relationship was noted for wheezing. An unusual feature of their analysis was to perform stratified analyses of parent-offspring association for asthma and "allergic" rhinitis, showing that specific associations for asthma and rhinitis independent of one another existed. Similarly log transformed total sIgE level (logIgE) in children was correlated with the parental levels, though variably (see Table). Parental logIgE levels were positively correlated with cigarette use (50% of fathers and 38% of mothers were smokers). When parental logIgE was adjusted for the presence of smoking, this removed any effects of sex. An individual's logIgE and skin test index were highly correlated (size not stated).
Table 22. Intraclass correlations for logIgE between parents and offspring [from Lebowitz et al 1984].
ParentChildrN
FatherSon0.09*115
Daughter0.2897
MotherSon0.30132
Daughter0.13*119

* P value > 0.05.

ANOVA analyses for various pairs of relatives were performed but not presented. It was noted that the most significant results were from the sib pairs. ANOVA with the children's logIgE as the dependent variable and age and smoking status entered as covariates confirmed that parental logIgE was a significant covariate of offspring logIgE, and that parental skin atopy was not a significant independent covariate. However, only 21% of the variance of the offspring logIgE was explained. With the child's skin test index as dependent variable, 26% of the variance was explainable using the parents logIgE and skin test results - with significant main effects for paternal skin atopy and maternal logIgE, and significant terms for interaction between skin reactivity and logIgE for both parents.

In the discussion, the authors comment that "It was not possible to determine the specific forms of genetic inherence [sic] in the manner of Rao et al [1980; see below] or Elston & Stewart [1971]". They note that the concordance among siblings might suggest either shared sibling environment or a genetic "factor of penetrance ... associated with polygenic factors", such as dominance or epistasis.

Meyers, Marsh and others [1986] on IgE

This group has performed a number of genetic studies on sIgE and skin atopy, including molecular genetic work described below; three were segregation analyses of sIgE in Amish and Mormon pedigrees. In another study [1986], they performed discriminant and regression analysis of the correlations observed between members of 42 nuclear families (N=278 individuals).

Families were randomly ascertained through a parent who was enrolled in an ongoing cohort study of Westinghouse Electric Corporation workers (N=2097), and then further selected for large size. Skin prick testing (9 common allergens), total sIgE levels, HLA phenotype and other blood groups were determined. In 230 cases, sIgE was determined on two occasions. An allergen index was calculated for the skin prick test results, based on the wheal size and importance of the allergen.

The log transformed total sIgE level (logIgE) was normally distributed within each sex, with the mean value for males was higher than that of females. The test-retest correlation for logIgE was 0.97. Because of the mean age of the subjects (28 years), there was no age related fall off in sIgE, save in skin test positive females (r=-0.44). Skin atopy (as defined by the Allergy Index) was present in 35% of individuals, in 42% of males, and 27% of females. The distribution of the 97 positive Allergy Index values was also normally distributed. In 13 families, no member exhibited skin atopy. The correlation between basal (ie out of pollen season) logIgE and Allergy Index was 0.48. For the genetic analysis, the logIgE was age-sex standardised, and intraclass correlations calculated for the various relationships (see Table).
Table 23. Familial correlations for logIgE and Allergy Index [from Meyers et al 1986].
FatherMotherOffspring
Father logIgE1.00
Mother logIgE0.33**1.00
Offspring logIgE0.330.290.43*
FatherMotherOffspring
Father AI1.00
Mother AI0.22**1.00
Offspring AI0.17**0.16**0.43*

* Correlation between siblings.
** P=0.05, all other correlations P<0.05.

The logIgE was positively correlated between different members of the family, including a marital correlation of marginal statistical significance. By contrast, only the sibling correlations achieved significance for the Allergy Index. A child in 39 families was then chosen as an index case, and a discriminant analysis performed to determine how accurately the index case's Allergy Index was predicted from his or her logIgE, and from the values from other members of the family. The presence or absence of skin atopy was correctly classified in 82% of cases by using the index logIgE. Classification using the siblings' mean Allergy Index was correct on 74% of occasions, and sibling and parental logIgE correct on 72%. Parental logIgE predicted correctly on only 54% of occasions overall. The same pattern of results was confirmed on stepwise multiple regression.

This study reports what must be regarded as preliminary results. Little discussion was made of the stronger correlations observed for siblings compared to the parent-offspring correlations, save to mention the possibility of sibling shared environment; no mention was made of genetic dominance as a possible explanation. The discriminant analysis was less informative but made use of the same information as would a comparable path analysis.

Meyers et al [1987] describe MVN variance components and mixed-model segregation analysis (PAP) of age-sex standardised logIgE among these subjects. The best fitting path model contained a parent-offspring correlation (modelled as an additive genetic component of variation, þ21=18.2), and a marital correlation (parental common environment, þ21=8.0). The narrow heritability was thus 31% (95%CI=7-55%). In the segregation analysis, the mixed model was superior to both the polygenic and single major gene models. Under this model, a recessive gene with increasing allele (A) frequency 0.97 (mean(AA)=98 IU/ml; mean(AB)=mean(BB)=3 IU/ml) explained 18% of the variance, and polygenes the remainding 19%.

A reanalysis of this data using a Class D regressive model [Meyers et al 1991] which included age, sex and allergy index as a covariate, reached essentially the same conclusion. The best fitting model was the high frequency autosomal recessive (single distribution with marital, parent-offspring and sib-sib correlations, LLik=-479.1; two distributions, free taus's, residual correlations, LLik=- 473.6; two distributions, Mendelian taus's, residual correlations, LLik=- 474.4). For this model, P(A)=0.99, mean(AA)=153, mean(AB)=mean(BB)=3. This work then supports a rare dominant decreasing gene.

In an earlier paper, Hasstedt, Meyers and Marsh [1983] reported the results of complex segregation analysis of total sIgE level in five pedigrees (N=316 individuals) ascertained through strong histories of familial breast cancer (including the family described by Gardner & Stephens 1950). The study's main aim was to examine the relationship between atopy and the putative protection it gives against cancer (see above). Serum for determination of total sIgE level was collected on two occasions. As usual age-sex standardised log transformed sIgE levels were used for analysis (logIgE). Complex segregation analysis was performed using PAP.

The mean sIgE levels of the cancer probands did not differ significantly from that of the other family members. The polygenic model fitted significantly better than the "sporadic" random environment model (see Table), giving a heritability estimate of 61% (95% confidence interval 40-82%). The major gene models were also significantly better than the sporadic model, but no particular mode of inheritance was preferred. The codominant model gave rise to a predicted gene frequency of 50%, midway between that posited in the dominant and recessive models - the authors suggest that this was a consequence of fitting these models to a polygenic trait. The fit of the mixed models was significantly better than that of the monogenic models, with polygenic heritability explaining a significant proportion of the variance. These were not significantly superior to the unadorned polygenic model.

In concluding, Hasstedt et al suggested that the differences in mechanism of inheritance found in the various studies in this area might be due to the timing of blood sampling. If sIgE level was determined on a sample collected during a pollen season, the values of atopic subjects would be higher than that of nonatopic relatives, leading to a bimodal commingled distribution, as detected by Gerrard et al [1978]. For the study under discussion, the samples were collected throughout the year, and unfortunately tended to occur at the same time of the year for members of the same family.

Table 24. Likelihood ratio X2 for genetic models considered by Hasstedt et al [1983].
ModelX2dfp
Sporadic0.00
Polygenic9.2410.00
Monogenic recessive6.7120.04
Monogenic dominant7.0120.03
Monogenic codominant8.7130.03
Mixed recessive9.8030.02
Mixed dominant10.0830.02
Mixed codominant10.0940.04

The paper by Meyers, Bias and Marsh [1982], describes complex segregation analysis of total sIgE level in 23 Pennsylvania Amish nuclear families. These were randomly selected. There was a total of 208 individuals and 14 of the families came from two extended kindreds and were so analysed. Individuals were designated as non, slightly, moderately or severely allergic using interview data. Skin prick testing was performed using 4 (mixed) extracts, and total sIgE level was determined on one occasion. The analysis was performed under the Elston and Stewart model using the program GENPED. As usual, the log transformed age- sex standardised total sIge level was used. Month of blood collection was examined and found not to contribute to variation in level.

There were 46 individuals (22%) with moderate or severe allergic symptoms. Only 32 subjects underwent skin atopy testing. Only 5 out of 11 "moderately" allergic subjects demonstrated any skin atopy. Though this differed significantly from the results for the nonsymptomatic subjects, predictive value was not high. This was thought to be due to a low prevalence of sensitisation to the usually important allergens that were used in testing; many allergic subjects reporting symptoms provoked by "barn dust" (?storage mites). Commingling analysis was performed; one distribution was found to be an adequate description of the frequency distribution of logIgE values. In the segregation analysis, the (random) environmental model was clearly rejected. The Mendelian models gave rise to overlapping phenotypic values. A good fit was found for the recessive (for high sIgE) model when compared to the equivalent unrestricted model, a poorer fit for the codominant model, and worse still for the dominant model (see Table). Some difficulties were met with parameters converging on their bounds. The authors then compared each dominance model (both Mendelian and unrestricted) with the unrestricted codominance model. In this case the codominant model was preferable.

The authors concluded that genetic factors were again seen to be important but that evidence for a major gene in their data was unsatisfactory.

Table 25. Likelihood ratio X2 for genetic models considered by Meyers et al [1982].
ModelX2dfp
Unrestricted (recessive)0
Environmental32.411,2*0.00
Recessive2.252,3*0.52
Unrestricted (dominant)0
Dominant10.2330.02
Unrestricted (codominant)0
Codominant8.842,3*0.03
Dominant9.2020.01
Recessive8.4720.01
* One or two parameters on bound.

Gerrard, Rao & Morton [1978] on IgE

These authors [1978] reported the results of path and complex segregation analyses (under the mixed model) of total serum IgE level in 173 Canadian nuclear families. Most (145) of these families were ascertained through a proband born at University Hospital, Saskatoon who had been under the care of a pediatrician, and were a fairly representative sample of the population. The original paper states that the remaining 28 families were recruited because of a high prevalence of atopic disease, but this is specifically corrected in a later reanalysis [Rao et al, 1980]. Serum was collected from the parents (N=336) and children (N=435) and total serum IgE, IgA, IgG and IgM levels were measured. The log- transformed sIgE was adjusted for age and sex. sIgG, sIgA and sIgM were used to create an environmental index - a procedure that might be of limited utility in view of the known genetic determination of these measures [e.g. Martin et al, 1982]. Busaniche et al [1985] report however that total sIgE and IgA, at least, were uncorrelated in 1030 individuals, though secretory IgA tended to be lower in young atopic individuals than controls. Six variables were entered into the analysis: the maternal, paternal and child's phenotype and environmental index (Table). Seven parameters were estimated for a submodel of the general nuclear family path model of Rao, Morton et al [1978] embodying an additive (multifactorial) genetic component, and a shared environmental component (E-E) with a unique marital environmental correlation, parent-offspring environmental correlation, and sibling environmental correlations. Separate maternal, paternal and child correlations to the environmental index were estimated.
Table 26. Observed correlations (x100) between parents and children for sIgE and environmental index [from Gerrard et al 1978].
F IgEM IgEC IgEF IndM IndC Ind
Father adjusted sIgE---
Mother adjusted sIgE08---
Child adjusted sIgE312226*
Father Index120410---
Mother Index-06241212---
Child Index1618233144---

* Correlation between siblings.

The general form was the preferred model (X2(9 df)=6.73, P=0.66); both the shared environment model (X2(4 df)=26.30, P<0.001) and pure genetic model (X2(1 df)=46.62) were rejected. This gave estimates for h2 of 42% (95% CI 31-54%) and via the environmental index for c2 of 7% (2-12%).

The authors then reported the results of the complex segregation analysis (using NUCLEAR). Log sIgE level was first normalised within generations and a power transformation applied to remove the remaining skewness. Two significant components were detected via commingling analysis during this second transformation. This was subsequently confirmed in Rao et al [1980] and Hasstedt et al [1985], and suggests the presence of a major gene. The segregation analysis suggested the presence of both a major gene (q=t=d=0, X2(3 df)=30.94) and polygenes (H=0, X2(1 df)=8.49) determining total sIgE. The major gene acted as an autosomal recessive (d=B=0, X2(2 df)=0.00) to increased levels of IgE, with an estimated gene frequency of 0.488. This would suggest that 24% of the population studied would be homozygotes, suggestively not dissimilar from the population prevalence of skin atopy. The rejected autosomal dominant hypothesis (X2(1 df)=12.69) of course gives the same number of "affected" heterozygotes and homozygotes. The model was further tested by dichotomising the population by sIgE level into low and high (normalised log(sIgE)>0.4 z units) and examining the results of the various mating types using the mixed model. Again, evidence for a major recessive gene was found.

The authors concluded that the shared environmental component detected in the path analysis may in fact have been a dominance component due to the autosomal recessive major gene (VD=13% and VA=13% as q=0.5). They expressed the pious hope that this gene would soon be identified (see below).

These data were partially reanalysed by Ott [1979], who used an EM algorithm based alternative numerical approach. He chose to analyse the power transformed IgE level assuming one component to the distribution and used an unconditional likelihood approach, in contrast to Gerrard et al, where likelihoods were conditioned on the parental phenotype. As a result, he found no evidence for segregation of a major gene. These methodological differences were pointed out by Rao et al [1980], who performed a further reanalysis using POINTER. This analysis supported Ott in finding no evidence for a major gene if a one component distribution and joint likelihoods for parents and children were assumed. Under the two component transformation, the authors' original findings were supported, as they also were under the conditional likelihood approach. The choice of transformation was reconfirmed with use of an improved version of SKUMIX; it must be noted that overall evidence of commingling apparently came from the children rather than the adults. However, the phenotypic distributions of parents and children were homogenous on testing. Blumenthal et al [1981] report an analysis of the same data using the Elston and Stewart model. The random environment model was again clearly rejected, but neither the recessive and dominant models were favoured. They comment that this method cannot of course differentiate between polygenic and monogenic models.

In Borecki et al [1985], information on atopic and other disease status from these 173 families was incorporated into the segregation analysis. Diagnoses were asthma, hayfever, eczema, recurrent bronchitis, allergic rhinitis, and urticaria; the criteria were symptom based. Using these criteria, 17% (30) families were nonatopic. The mean total sIgE level was elevated among symptomatic subjects, but was normal in many subjects with allergic rhinitis and recurrent bronchitis. Atopic affection status was therefore divided into three levels:

  1. Asthma or eczema (prevalence 21%)
  2. Asthma, eczema, hayfever, or recurrent bronchitis (32%)
  3. Asthma, eczema, hayfever, recurrent bronchitis, allergic rhinitis or urticaria (45%)

Affection status and transformed IgE were positively correlated: 0.20 for level 1 and 2, and 0.17 for level 3, and cross trait parent- offspring correlations were present (see Table). Segregation analysis using the unified mixed model for a threshold trait was then performed (using POINTER).

Univariate analyses were first presented for the three affection states in turn. Although in each case, the nongenetic model was rejected, the major gene model could not be differentiated from the multifactorial model. A bivariate model entering normalised IgE (as earlier) and affection status was fitted, again separately to the three different affection levels. This model involved an extension of the usual mixed model (where x=normalised IgE and y=liability to allergy):

Equations not shown

Four parameters for y additional to those in the univariate model are estimated: dy=dominance; ty=displacement; Hy and Ey the heritability and environmental variances specific to y. The parameter w is the specific residual deviation in liability; W is the corresponding variance component and is completely determined as the liability scale is standardised. For model testing, dx was set to zero (see Table). For liability level 1, in the general hypothesis, Hy was estimated as zero, implying that the multifactorial component was unimportant to the determination of liability, in contradistinction to IgE level. A major gene with codominant inheritance was the preferred model, with no environmental variance specific to liability. For liability level 2, however, the preferred model was the codominant model, and for level 3 (the least specific), a dominant inheritance provided the best fit. Parental
Table 27. Parent-offspring correlations for presence of atopy (level 3) and sIgE [from Borecki et al 1985].
Offspring
sIgEAtopy
ParentsIgE0.250.17
Atopy0.090.22

Table 28. Bivariate segregation analysis for sIgE and level 1 liability to allergy [from Borecki et al 1985].
HypothesisLR X2dfp
General (Hy estimate=0, dx set=0)0.0
No association IgE and liability25.640.00
No shared genetic component16.5830.00
No major gene9.2820.01
No specific environment ie Ey=00.6810.41
Recessive major gene2.2710.13
Additive (codominant) major gene0.7410.39
Dominant major gene12.2210.00

To check for possible confounding due to age on liability, the analyses were repeated using age specific prevalences and correlations for level 3 atopy. Fit was actually improved, and no significant differences in the parameter estimates or hypothesis tests were detected. The authors interpreted their findings as supporting the hypothesis of an IgE regulator gene, where RR and Rr lead to lower sIgE and rr to higher levels, but where heterozygotes have increased risk of exhibiting atopic disease, despite a normal total sIgE. The population frequency of allele r (as in the first study) is between 42-48%.

A preliminary study [Gerrard et al 1976] of this material is interesting because it examines specific genetic influences. This included 176 families (rather than 173). They include a delightful table of X2 values for the 2x2 tables formed to examine within-child and parent-offspring concordance within and across asthma, hayfever, "recurrent" rhinitis, recurrent bronchitis, eczema and urticaria. These represent data from 475 relative pairs and are not corrected for nonindependence. The same trait parent-offspring correlations were higher than the cross-trait correlations. I have presented these correlations as binary correlations (Table).

Table 29. Parent-offspring binary correlation coefficients from Gerrard et al [1976].
Offspring
ParentAsthmaHayfeverRec. RhinitisRec. BronchitisEczema
Asthma0.29
Hayfever0.160.18
Rec. rhinitis0.070.030.14
Rec. bronchitis0.210.030.020.22
Eczema0.070.010.050.050.15

McGue et al [1989] on sIgE

McGue, Gerrard, Lebowitz and Rao [1989] performed commingling analysis on immunoglobulin levels in sera from both the Saskatoon and Tucson studies examined above. This gave a total of 1745 North American adults and children. Total sIgA, D, E, G and M levels were age and sex adjusted, and in the case of IgE further adjusted for month of blood collection in (skin) atopic individuals only. The results were power transformed and commingling analysis performed. In the Canadian sample, the presence of commingling was detected for the children's sIgE levels, and for IgA, IgG, IgD in both adults and children. No commingling was detected in the Tucson results, with the exception of children's sIgE. These results accord with the findings that heritability of sIgE has been found to be higher in children than adults in a number of studies. For the Canadian data, a two component model was preferred (1 v 2 components, X2(2 df)=7.3, P=0.03; 2 v 3 components, X2(2 df)=0.5, P=0.78) with a gene frequency of 0.51 for the recessive gene associated with a high sIgE level. In the Tucson data, the three component model was favoured (1 v 2 components, X2(2 df)=6.9, P=0.03; 2 v 3 components, X2(2 df)=6.8, P=0.03), corresponding to a gene frequency of 0.38. The deviation associated with the "high" genotype was 1.65 for the Saskatoon, and 2.63 (with d=0.45) for the Tucson children.

Blumenthal et al [1981] on IgE

Blumenthal, Namboodiri, Mendell, Gleich, Elston and Yunis [1981] examined three large pedigrees (N=184 individuals) by complex segregation analysis using the method of Elston and Stewart. The families were ascertained through a history of multiple members suffering ragweed pollinosis. Total sIgE level (serum collected out of pollen season) was log transformed and age-sex standardised (either before segregation analysis, or simultaneously). A mixture of two lognormal distributions was found to fit the logIgE distribution, and was marginally better than using the conservative one distribution model (X2(2 df)=5.76, p=0.056).

The random environment model was rejected (X2(3 df)=27.03). The polygenic model gave an estimated heritability of 50%. The favoured mode of transmission was an autosomal dominant gene for high IgE. An approximation to the "mixed model" estimated the additional polygenic heritability at 10%, which was not significant, but the authors note that the procedure used favoured the major gene (in variance terms). The displacement of the means was 2.0 SD units. There was evidence of heterogeneity between the families. A borderline significant difference in sIgE levels in the different pedigrees was found to be due to a lower mean sIgE level in the females of one pedigree. Homogeneity þ2 testing of the dominant and recessive hypotheses demonstrated that the pedigrees differed from one another (dominance X2(18 df)=30.05; recessivity X2(18 df)=42.09). Separate analyses of each family were performed. In the first pedigree, the preferred model was clearly autosomal dominant. The second family favoured either hypothesis, while the third favoured the recessive inheritance.

The authors concluded that "[t]here is a strong possibility of either multiple alleles at a single locus or involvement of more than one major locus". This is quite possible in this case as ascertainment was through ragweed allergy. As discussed earlier, ragweed pollinosis is the only specific hypersensitivity with a HLA association, confirmed by an RFLP linkage study. These particular families (or the first and second families) may represent epistasis between a general atopy gene(s) and the Amb 1 V sensitivity gene.

This group has also reported a path analysis of basophil histamine releasability [Roitman-Johnson & Blumenthal 1988; Blumenthal 1993]. In this assay, anti-IgE antibodies or other agents are titrated against histamine release from blood; alterations (increases or decreases depending on the agent) have been reported in atopy [Marone et al 1986]. In ten nuclear families ascertained through a ragweed-sensitised proband, there was a sib-sib correlation of 0.85, with a lower parent-offspring correlations. The authors estimated the heritability of anti-IgE provoked histamine releasability at "55-70%" [Blumenthal 1993].

Parks et al [1990] on sIgE and atopy

This recent study [1990] included Rice and DC Rao as coauthors. It examines total sIgE level and atopic symptoms in the nuclear families of 42 asthmatic proband ascertained through an outpatient clinic in Madras (India). As in Borecki et al [1985, see above], categories of reported symptoms were used to define two affection classes, one containing the core syndromes of asthma, allergic rhinitis and atopic eczema, the other including these as well as nonasthmatic wheezing and questionable allergy symptoms. Since ascertainment was nonrandom, and the appropriate values were not known, the population prevalence of these two entities was estimated from that in the relatives of the probands, including and excluding the probands, giving ranges of 12-27% for the prevalence of the core syndromes and 15-31% for the broad category. Taking the authors' "reasonable estimates" of 15% and 20% for the prevalences leads one to assume that atopic disease is less common in this population.

The mean total sIgE level was 637 U/ml, higher than that generally reported in Western studies. The log transformed age- sex standardised total sIgE level was used for the segregation analysis. Commingling analysis (SKUMIX) found two (log) normal distributions to best fit the logIgE frequency distribution, so segregation analysis using POINTER was performed on the un(- power-)transformed variable. The clinical syndromes were included using the threshold model, with liability to affection (y) a linear function of adjusted sIgE (x; y=x+w; E(x)=0, Var(w)=W).

No evidence for a major gene was found (the gene frequency converging on zero on each attempt). The familiality (since no environmental index was incorporated) was estimated at 0.2, quite low when compared to other studies. The correlation between liability and logIgE was 0.4, as seen above. The authors considered that these results might reflect a higher parasite load in this population, so that heritability of sIgE is proportionately decreased. This leaves the task of explaining why the prevalence of atopy was probably lower than that seen in Western studies. The authors invoked the hypothesis that parasitosis protects against atopic illness, but as discussed earlier, the relationship might run in the opposite direction, so an alternative suggestion is that exposure to major allergens is less prevalent in Madras.

Silverman et al [1990]

This another study [1990] of sIgE including DC Rao among the authors. As noted in Section 3.8, A1AT deficiency states are associated with increased asthma or asthma-like symptoms. Therefore these authors performed a measured genotype analysis using 52 PiZZ (and Z-) probands and their immediate families - 44 segregating nuclear families from 29 pedigrees (total of 167 subjects). Path and complex segregation analyses were performed on log total sIgE level, and total and oxidised alpha-1-AT level.

Since Pi type was found to be uncorrelated with sIgE level, analyses were performed with and without ascertainment correction. In the path analyses of sIgE, a marital correlation of 0.27 was observed, with the overall heritability being 0.27 under additive genetic transmission (with sibship and marital genotypic correlations modelled). The heritability of sIgE was 0.64 in children and 0.47 in adults under the no-major-gene POINTER model by contrast (conditioning on parents and probands; versus the no transmission model X2(df=2)=16.8, P=0.00; the data for conditioned on parents alone not shown). The major gene model fitted equally well (versus the no transmission model X2(3 df)=18.0, P=0.00), as of course did the mixed model (X2(5 df)=18.4, P=0.00). The heritability estimate agrees well with those above. The sample size seems to have been insufficient for testing of major gene models.

Dizier et al [1994]

A complex segregation analysis of total sIgE in 234 Busselton nuclear families was reported by these authors [Dizier et al 1994]. Fitting the Class D regressive logistic model detected a sex-limited common (q=70%) major autosomal recessive gene explaining 30-50% of trait variance (depending on sex and generation), with residual familial correlations (polygenic or shared environmental r=0.25). The difference between genotypic means was equivalent to a 3.7 fold difference in (raw) IgE levels for females, and a 7.4 fold difference for males. Approximately, the population of male children was modelled to be a 50:50 mixture of "lows" (geometric mean 30 IU/ml, interquartile range 3-260) and "highs" (geometric mean 220 IU/ml, IQR 25-1900).

When atopy (any positive skin prick test or positive RAST) was added as a covariate, "the residual familial correlations [were] no longer significant". The authors interpreted these findings as suggesting the major gene controls basal sIgE level, but that there are additional familial factors (probably genes) acting on specific sensitization.

The ideal approach to this data of course would be a multivariate one, as was seen earlier for Borecki et al. Examining the familial correlations suggests that the measures of specific sensitisation explain approximately one-third of the familial aggregation of total sIgE (Table). However, we do not have the reverse analysis for specific sensitisation.

Table 30. Familial correlations for age-sex-generation adjusted log total sIgE from Dizier et al [1994]. The spouse correlation was not significantly different from zero.
Covariate-adjustmentFather-ChildMother-ChildSib-Sib
Nil (age-sex-gen)0.280.260.31
Specific IgE or SPT0.180.210.24

Lawrence et al [1994]

The Southampton group (Steven Holgate on the asthma side, and Newton Morton on genetics) reported a path and complex segregation analysis of atopy and asthma. The 131 nuclear families (631 completely phenotyped individuals) were selected only to contain sibships of three or more, and can be regarded as a random sample with respect to asthma.

Clinical phenotypes were defined using a questionnaire based on that of the IUATLD supplemented by a video. Total sIgE determinations, skin prick tests to 11 common allergens, and Yan histamine challenge were performed.

Atopy was defined as the factor score for the first principal component of log total sIgE, AUC of dose-response curve for histamine, the product of the diameters of the SPTs, and summed questionnaire scores for wheeze, hayfever and eczema; asthma was defined similarly using BR and wheeze.

Simplified path models were fitted to the data (Table). The highest heritability estimated was for log sIgE at 61% (95%CI=50-72%). That for BR was 27% (14-40%), and for product of SPTs 34% (21-47%). Commingling analysis was suggestive of a mixture of distributions.

Two types of segregation analysis were performed following inverse normal transformation (POINTER) and subsequent polychotomization (COMDS). The mixed model in POINTER supported a polygenic model for sIgE, and a SML for BR and SPT with zero residual heritability. The authors were sceptical of these latter findings, especially because SPT has a number of nonreactive individuals (zero scores), forcing a SML model. Similar two locus models using COMDS, setting the modifier locus parameters to simulate polygenic background (dominance dm=0.5, gene frequency qm=0.5) suggested common major genes of small effect, though giving rise to different segregation parameter estimates than the equivalent POINTER analyses. For example, POINTER's general model for transformed total sIgE gave a recessive model with gene frequency (q) of 0.62, residual heritability (h2) of 38%, and likelihood X2 (3 df, compared to the polygenic model) of 4.0; COMDS gave an additive model with q=0.22, (pseudo) h2=12%, and LR X2(3 df)=12.9. For the atopy trait, POINTER gave a recessive gene, q=0.55, h2=40%, LR X2(3 df)=2.5; COMDS an additive gene with q=0.25, h2=10% and X2(3 df)=11.9. To my mind, the segregation results seem a little inconsistent, especially across the two programs used.

Table 31. Selected familial correlations from Lawrence et al [1994].
TraitSpousal (N=130)Father-Child (N=303)Mother-Child (N=297)Sib-Sib (N=334)
log sIgE0.080.360.280.38
BHR0.050.150.080.20
SPTs0.060.210.140.20

Martinez et al on sIgE [1994]

Among 291 nuclear families ascertained through the Tucson Children's Respiratory Study (N=1077) described in this paper, 50 were Hispanic in origin, allowing testing for ethnic (genetic) heterogeneity in a segregation analysis of sIgE level [Martinez et al 1994]. The parent-offspring correlations for log sIgE were a little lower than those seen in other studies (sib-sib 0.31 +/- 0.06, parent-offspring 0.18 +/- 0.02, spousal 0.05 +/- 0.06). Using SAGE REGC with a Box-Cox "link" function to simultanously estimate age-sex standardised Z-scores, the authors found evidence for ousiotypes (that is a mixture of distributions - X2(7 df)=88.0), with close to Mendelian tau's (tau(AA)=1.00, tau(AB)=0.44, tau(BB)=0.00), rejecting the environmental transmission model (X2(4 df)=85.5). The codominant model fitted better than the dominant or recessive, and there was marginal support for "polygenic" residual correlations. As usual, the gene frequency was high, estimated at 0.34 for the increasing allele (A), which acted in a near additive fashion (on the z-transformed scale). No ethnic heterogeneity was detected.

Sampogna et al on sIgE [2000]

The EGEA study recruited 335 French nuclear families (1414 individuals) through an asthmatic proband. In this paper, various ascertainment corrections are made to a Class D regressive segregation analysis of serum IgE level. The REGRESS package was used. The familial correlations for sIgE were consistent with a heritability of about 40% (marital correlation approx 0; sib-sib 0.21; father-child 0.20; mother-child 0.30). The analysis ignoring the ascertainment mechanism failed to detect a major gene, though the presence of multiple ousiotypes was supported. Both exclusion of probands and conditioning on parental phenotype, and regressive adjustment of sIgE according to relationship to the proband, led to support for a common (P(A)=0.64) dominant major gene model explaining 14-17% of the variance (Table).

Table 32. Segregation analysis for sIgE conditioning on parental sIgE level and excluding asthmatic probands [from Sampogna et al 2000].
Hypothesis-2*LLNo. parametersAIC
No major gene, no familial r494.72498.7
No major gene, 3 familial r421.65431.6
Major gene, 3 familial r387.58403.5
Dom gene, 1 familial r389.05399.0
2 means, env trans, 1 familial r394.66406.6
2 means, free taus, 1 familial r388.08404.0

Palmer et al on sIgE, BHR and eosinophil count [2000]

These papers present results of segregation analysis performed using BUGS. of the Busselton "asthma" families (they are unascertained for that trait). SML models were preferred for each univariate analysis, and pleiotropy was looked for by linkage analysis using the imputed trait locus genotypes using VITESSE.

The sIgE locus had an increasing allele freuqncy of 69%, and acted in a recessive fashion (adjusted log[sIgE] genotype means 0.28, 0, 2.01). A logistic regression predicted the increasing homozygote to increase risk of asthma 1.45-fold. It explained 15% of the total phenotypic variance of tIgE (and one-third of the total heritability) The eosinophilia locus similarly explained 10%, and the BHR locus 27% of the population variance. The imputed loci were all unlinked.

Mathias et al [2001] describe a genome scan for loci controlling tIgE, but mention in passing heritability estimates based on the family correlations in the CSGA study. For log-transformed tIgE, these were 60% for black Americans (107 families), 56% for white Americans (129 families), and 44% for Hispanic Americans (32 families).

Xu et al on sIgE [2000]

A two-locus segregation analysis of tIgE using PAP was reported by Xu et al [2000]. The Groningen sample of families was extended now to 200, 66 of which were three generational. The overall heritability of log-transformed tIgE was 55%, and a SML segregation analysis was consistent with a common recessive locus (increasing allele frequency 57%) with genotypic age-sex adjusted geometric means of 30 IU/ml and 210 IU/ml. The two major locus model gave a better fit to the data, and could be simplified to a four-mean model (with A recessive, and B dominant-negative to A):

P(A)=0.45, P(B)=0.2

AAAaaa
BB1.360.141.36
Bb1.361.362.45
bb1.531.532.45

The heterosis (BBAA > BBAa < BBaa) might be due to a floor effect in sIgE level, exaggerated by the log transformation of the trait.

Animal pedigrees

Dogs and ponies spontaneously exhibit diseases similar to human atopy. Dog atopy is usually seasonal, and is characterised by a pruritic face, axillae and paws [Schwartzman et al 1971]. A Prausnitz-Kustner reaction can be elicited, and hyposensitisation. Sensitization is to a range of allergens similar to those in human disease, prodominantly house dust mite, mould, fleas, pollen (Zur et al 2002; Youn et al 2002). The prevalence in the general population is 0.5--1.0%.

In breeding experiments using atopic dogs, Schwartzman et al found atopy (including leucocyte histamine releasability) to be strongly heritable, and pedigrees were most consistent with an autosomal dominant inheritance. Particular breeds also exhibited higher prevalence of the trait.

Shaw et al [2004] report on clinical atopic dermatitis in Labrador and Golden Retriever pedigrees ascertained through 13 probands (N=429). The heritability was estimated at 47%.