- Email: [email protected]
Familial Aggregation of Airway Responsiveness: A Community-based Study
Familial Aggregation of Airway Responsiveness: A Community-based Study KE HAO, SCD, CHANGZHONG CHEN, MD, BINYAN WANG, MD, PHD, JIANHUA YANG, MD, ZHIAN FANG, MD, AND XIPING XU, MD, PHD
PURPOSE: We investigated the familial aggregation of airway hyper-responsiveness (AHR) to methacoline among randomly chosen families in a rural community in Anqing, China. METHODS: Airway responsiveness (AR) to methacoline and related risk factors were assessed in each subject. We first modeled the within family correlation in AR and demonstrated the familial aggregation of this trait. Furthermore, we examined the effect size (e.g., odds ratio, OR) of this correlation in a ‘‘subsequent offspring model.’’ RESULTS: The correlation coefficient is significantly positive for parent-offspring and offspring-offspring pairs, but not significant in father-mother pairs, suggesting a genetic component. The strength of the relationships is in the order of father-offspring ! mother-offspring ! offspring-offspring. The OR of a positive AHR test for subsequent offspring who had mothers and an eldest sibling with positive AHR is 4.12 (95% CI, 1.72–9.87), compared with subsequent offspring whose mother and eldest sibling were negative in the test. CONCLUSION: Our study supports a familial clustering of AHR in a Chinese population, which points to a role for genetic factors. Ann Epidemiol 2005;15:737–743. Ó 2005 Elsevier Inc. All rights reserved. KEY WORDS:
Familial Aggregation, Genetic Epidemiology, Asthma, Airway Responsiveness.
INTRODUCTION Many genetic and environmental factors could be involved in the development of asthma (1), a complex disease with the following characteristics: 1) airway obstruction that is reversible, either spontaneously or with treatment; 2) airway inflammation; and 3) airway hyperresponsiveness (AHR) to a variety of stimuli (2, 3). Like other complex disorders, asthma does not follow a simple Mendelian inheritance pattern (4). In mapping the asthma susceptibility gene, several intermediate phenotypes are used, including AHR (5, 6), total or specific immuno-globulin E (IgE) (7, 8), and blood eosinophil count (EOS) (9, 10). AHR, which has been considered a central feature of the definition of asthma, is usually regarded as an abnormal response of the lower respiratory tract to a number of nonsensitizing bronchoconstrictive stimuli (11–14). It is well recognized that there are significant inter-individual variations in airway responsiveness (AR) (15, 16) and AR is closely associated with asthma (17). Before focusing on the
From the Program for Population Genetics, Harvard School of Public Health, Boston, MA (K.H., C.C., B.W., X.X.); and Institute for Biomedical Research, Anhui Medical University, Hefei, China (J.Y., Z.F., X.X.). Address correspondence and reprint requests to: Xiping Xu, M.D., Ph.D., Program for Population Genetics, Harvard School of Public Health, 665 Huntington Ave. FXB-101, Boston, MA 02115. E-mail: xu@hsph. harvard.edu Received May 6, 2004; accepted February 1, 2005. Ó 2005 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010
goal of locating susceptibility genes controlling asthma, a critical need to be addressed is, ‘‘does the disease cluster in families?’’ Using regressive logistic models, Jenkins et al. (18) showed positive familial aggregation of self-reported or clinic diagnosed asthma, and Palmer et al. (19) estimated the narrow sense heritability of lung function and doseresponsive slope to methacholine (MTCH) in the Busselton population, a population with a high asthma prevalence (20.2%). Xu et al. (20) found significant familial aggregation of lung function parameters (force vital capacity, FVC and forced expiratory volume in 1st second, FEV1) among asthma index and randomly chosen families. The familial aggregation of AHR, an important clinical characteristic of asthma, has not been well studied in a general population. To address this issue, we randomly selected 180 nuclear families from Anqing, China, and collected comprehensive phenotypic information. Compared with ascertained asthma families, these randomly selected families are more suitable for determining the familial inheritance patterns of quantitative intermediate phenotypes. We first modeled the within family correlation to examine the familial aggregation. Then we studied the relationship of AHR between subsequent offspring and their parents and elder sibling. This study helps clarify our understanding of the genetic epidemiology of AHR, and it may provide preliminary evidence for mapping genes underlying the intermediate phenotypes for asthma. 1047-2797/05/$–see front matter doi:10.1016/j.annepidem.2005.02.002
738
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
MATERIALS AND METHODS Study Subjects Anqing city is located in the central part of China with a population of 5.8 million in 1990. We chose Anqing as the study site for the following reasons: 1) the inhabitants are relatively homogeneous with respect to ethnicity, environment, occupation, and diet; 2) the stable population has existed for thousands of years; and 3) a uniform medical care service network has been established for more than 25 years, which offers a unique opportunity for efficient identification of study families. We selected families from the general population according to 1992 census records by a two-stage random sampling technique. At the first stage, we randomly selected villages from Anqing, and at the second stage, we randomly sampled one household in each selected village. Our inclusion criteria were: 1) a family size of at least four; 2) the availability of both parents; 3) two or more siblings in the family; and 4) at least 6 years old for the youngest offspring. Only the families that provided written informed consents were enrolled into the study. A detailed description of the study design and sample selection procedure has been presented elsewhere (20). Field Study Procedure Subject enrollment started in February 1995, after obtaining local China IRB approval. Brigham and Women’s Hospital IRB approval was received in September 1995. The field study was conducted by faculty members from Anhui Medical University, focusing on asthma and its intermediate phenotypes. The following procedures were carried out when the study subjects were invited to the research center: 1) standardized questionnaire (modified AST-DLD) assessing the respiratory history and symptoms, occupational and smoking histories, home environment, family history of asthma, and other chronic or genetic diseases; 2) pulmonary function test (spirometry); 3) methacholine challenge test for all subjects with FEV1 values O 60% of predicted value; and 4) bronchodilator test. In addition, weight and height were measured by standard methods after the subjects removed their shoes and outerwear. A detailed description of the study procedures has been presented elsewhere (20, 21). Measurement of Airway Hyper-responsiveness If baseline FEV1 was O 60% of predicted value, the airway challenge test was performed. A standard protocol for methacholine challenge test was developed based on a modified Chai protocol (22). Briefly, there were up to five steps to accomplish the methacholine challenge test with the following breath/dose schedules: saline solution, 5 mg/mL (one breath), 5 mg/mL (four breaths), 25 mg/mL (one breath), 25 mg/mL (four breaths). At each dose, two
AEP Vol. 15, No. 10 November 2005: 737–743
additional satisfactory spirometry maneuvers were obtained. The methacholine challenge test was continued until either a 20% drop of FEV1 from baseline was observed or the highest dose of the methacholine was reached, and then a following bronchodilator test was performed (22). We analyzed the dose–response relationship between methacholine dosage and FEV1 response of each subject in both continuous and binary forms. The continuous outcome is defined as the DFEV1/dose, where DFEV1 was defined as the drop FEV1 after the last methacholine challenge compared with the saline baseline, and dose was defined as the cumulative dosage methacholine administered. This variable was named as dose response slope (DRS) by convention. The binary outcome was defined as whether a subject showed a larger than or equal to 20% drop of FEV1 in the whole process, and was named PD20B. Statistical Analyses The families were randomly selected, and represented an unbiased and random sample of the general population. The DRS is strongly skewed to the right, and can be slightly negative, which means that FEV1 may increase after methacholine inhalation. This could be attributable to the biological phenomenon that the FEV1 bears certain degree of variation from time to time or from test to test. For example, the second measurement could give a slightly larger value than the first measurement, and resulted in a negative DRS. Another possibility is the learning effect that the participant performed better in the later maneuvers than the earlier ones. We somewhat controlled this confounder by letting the participants practice the spirometry a few times prior to the formal tests. A small number (1 unit) was added to all the DRS to make them all positive, and a log transformation is applied (Fig. 1). To test the impact of a few outliers, we conducted analysis by removing or including subjects outside the range defined by MeanDRS G 3SDDRS, and obtained similar results (data not shown). In this article we only present the analysis with all data points. Trait correlation among family members indicates the familial clustering. We modeled this correction using four parameters r1, r2, r3, and r4 standing for the father-mother, father-offspring, mother-offspring, and offspring-offspring correlation coefficients, respectively. In fitting the random effect model using a statistical software package (SAS, version 6.2), we specified an unstructured covariance matrix in order to estimate all these parameters. Y ¼ XbCZbC3; where Y denotes the outcome variable vector and X denotes the fixed effect variable matrix including all the potential
AEP Vol. 15, No. 10 November 2005: 737–743
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
739
FIGURE 1. The distribution of DRS and its log-transformation. Upper: The distribution of DRS. Lower: The distribution of Log(DRS). X-axis denotes the scale of the variable, and y-axis denotes the number of observation.
confounders including age, height, weight, gender, and smoking status. The random effect variable Z is a 775 ! 3 matrix (775 individuals in analysis). For any father, Zi has the value of [1,0,0]; for any mother, Zi has the value of [0,1,0]; and for any offspring, Zi has the value of [0,0,1]. The choice of Z specifies the structure of the within family correlation matrix. b and b are the fixed-effects and randomeffects parameters. The random effect model assumed b and 3 followed independent normal distributions.
E
b 0 ¼ 3 0
Var
b G0 ¼ 3 0R
predict the outcome of the subsequent offspring. The prediction value of family member’s AHR can also be considered as a measurement of the familial aggregation, and the odds ratio (OR) reflects the effect size. We fitted a ‘‘subsequent offspring model,’’ where the log(DRS) or PD20B of subsequent offspring were treated as dependent variable, and the value of his parents and the 1st offspring entered the model as covariates. A small number of families contributed two subsequent offspring, whose DRS or PD20B were hence not independent. Therefore, we applied a generalized estimation equation (GEE) with ‘‘exchangeable’’ correlation structure to adjust the standard error. RESULTS
The decomposition of V (variance of Y) was V Z ZGZ’ C R. We estimated G and R using restricted/ residual maximum likelihood (REML) (23), implemented in SAS PROC MIXED. The log-likelihood 1 1 lðG; RÞ ¼ log jVj log jX0 V1 Xj 2 2 1 0 1 np logð2pÞ; rV r 2 2 where r Z Y X(X#V1X)1X’V1Y and p is the rank of X, was maximized using the Newton-Raphson algorithm, and G and R were therefore estimated in the procedure (24). Upon detected significant familial aggregation, we were interested in using parents and/or the 1st offspring’s AHR to
The study included 180 families with complete data on pulmonary function and major covariates. The clinical characteristics of the study population are presented in Table 1. On average, fathers were 3.4 years older than mothers, and were more likely to be smokers than mothers and their offspring. The prevalence of clinically diagnosed asthma was low in this cohort (2.1 %). The average DRS and the percentage of positive methacoline test results (PD20B Z 1) were nearly the same among fathers, mothers, and children. First, we tested familial aggregation using a random effect model estimating the correlation coefficients for DRS among family members (Table 2). Fathers and mothers were slightly positively correlated but not significantly (cor Z 0.06), which might due to shared
740
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
AEP Vol. 15, No. 10 November 2005: 737–743
TABLE 1. Clinical characteristics of the 180 random families Mean (SD) or Number (percentage)
Father (n Z 180)
Mother (n Z 180)
First offspring (n Z 180)
Subsequent offspring (n Z 235)
44.4 (7.6) 1.65 (0.06) 57.4 (7.1) 0% 4.38 (0.65) 3.34 (0.60) 14 (7.7%) 142 (78.9%) 7 (3.9%) 1.01 (0.93) 48 (26.7%)
40.9 (7.8) 1.54 (0.05) 51.7 (6.7) 100% 3.37 (0.51) 2.66 (0.45) 0 3 (1.7%) 1 (0.5%) 0.93 (0.93) 44 (24.4%)
17.9 (7.0) 1.52 (0.13) 44.7 (12.2) 48.3% 3.39 (0.99) 2.90 (0.80) 2 (1.1%) 20 (11.1%) 4 (2.2%) 0.92 (0.95) 50 (27.7%)
15.6 (7.2) 1.44 (0.17) 38.5 (14.1) 47.2% 2.91 (1.09) 2.49 (0.88) 0 13 (5.5%) 4 (1.7%) 1.00 (1.02) 72 (30.6%)
Age, y Height, m Weight, kg Female, % Baseline FVC*, L Baseline FEV1**, L Former smoker Current smoker Diagnosed asthma, % Log (DRSx C1) PD20Bxx, %
*Forced vital capacity. **Forced expiratory volume in first second. x Methacoline dose–response slope. xx PD20B indicates that the individual experienced a R 20% drop of FEV1 during the methacoline challenge test.
environment factors. The correlation between fatheroffspring (cor Z 0.14, p ! 0.02) and mother-offspring (cor Z 0.14, p ! 0.02) pairs was comparable in our study. The highly significant positive correlation between siblings (cor Z 0.21, p ! 0.002) was observed. Second, a linear GEE regression model was fitted to test the relation of trait values between subsequent offspring and their parents and eldest available brother or sister. We found that the strength of the relation to subsequent offspring was in the same order as we observed in the random effect model: father ! mother ! eldest available offspring. The regression coefficient (b) of DRS between subsequent offspring and their eldest sibling was statistically significant in the models, although the magnitude of b was somewhat reduced in multivariate models due to the correlation of DRS among family members (Table 3). To diagnose or classify asthmatic status, the binary form of AHR is widely used and may be more meaningful in clinical practice. We applied logistic regression models and found the PD20B of subsequent offspring is significantly associated with that of their mother (OR Z 2.81; 95% CI, 1.39–5.70) and their eldest available siblings (OR Z 2.08; 95% CI, 1.06–4.07). The effect of the eldest available offspring was somewhat attenuated and becomes nonsignificant in the multivariate model (Table 4). We then TABLE 2. Correlation coefficients among family members (A family of two parents and three offspring) Correlation
Father
Mother
Offspring 1
Father Mother Offspring 1 Offspring 2 Offspring 3
1 0.063 0.138* 0.138* 0.138*
1 0.142* 0.142* 0.142*
1 0.206** 0.206**
*Statistically significant at 0.05 level. **Statistically significant at 0.01 level.
Offspring 2
1 0.206**
Offspring 3
1
introduced interaction terms into the model to further capture the familial aggregation pattern, and detected a significant interaction between the trait values of mother and those of her first available offspring (Table 5 and Fig. 2). The OR of having a positive AHR result for subsequent offspring who had a mother and elder sibling with positive AHR results was 4.12 (95% CI, 1.72–9.87). DISCUSSION In our study, correlation coefficients of family-relative pairs were significantly greater than zero, indicating a pattern of familial aggregation, which can be attributed to the sharing of genetic or environmental factors or both. The correlation was stronger in offspring-offspring and parent-offspring pairs than in parent-parent pairs, suggesting that genetic components may play a role in AHR. TABLE 3. Linear GEE regression analysis of log(DRS) of subsequent offspring in relation to their family members Adjustedx
Crude Model*
b
95% CI**
b
95%CI
Univariate Father Mother 1st sib
0.05 0.08 0.22
(0.08, 0.18) (0.02, 0.18) (0.09, 0.35)
0.08 0.15 0.20
(0.05, 0.22) (0.03, 0.26) (0.05, 0.34)
Multivariate Father Mother 1st sib
0.02 0.04 0.21
(0.10, 0.14) (0.04, 0.14) (0.07, 0.34)
0.05 0.10 0.14
(0.08, 0.18) (0.01, 0.21) (0.01, 0.27)
*Model: LogðDRSSubsequent Offspring Þ ¼ b1 XCb2 ZC3 Z is the matrix of environmental factors. In univariate model: X is the vector of [Log(DRSFather)i or Log(DRSmother)i or Log(DRS1st Offspring)i] In multivariate model: X is the matrix of [Log(DRSFather)i, Log(DRSmother)i, Log(DRS1st Offspring)i] **Standard error is corrected by GEE model. x Adjusted for age, sex, height, weight, and smoking status.
AEP Vol. 15, No. 10 November 2005: 737–743
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
741
TABLE 4. Logistic GEE regression analysis of PD20B* of subsequent offspring in relation to their family members Adjustedx
Crude Model**
OR
95%CIxx
OR
95%CI
Univariate Father Mother 1st sib
1.44 2.04 2.20
(0.74, 2.39) (1.08, 3.86) (1.16, 4.13)
1.62 2.81 2.08
(0.81, 2.73) (1.39, 5.70) (1.06, 4.07)
Multivariate Father Mother 1st sib
1.19 1.75 1.92
(0.59, 2.38) (0.90, 3.38) (0.98, 1.92)
1.37 2.44 1.68
(0.67, 2.83) (1.19, 4.99) (0.83, 3.40)
*PD20B Z 1 indicates that the individual experienced a R 20% drop of FEV1 during the methacoline challenge test. **Model: LogitðPD20 BSubsequent Offspring Þ ¼ b1 XCb2 ZC3 Z is the matrix of environmental factors. In univariate model: X is [PD20B Father or PD20B mother or PD20B 1st Offspring] In multivariate model: X is the matrix of [PD20B Father, PD20B mother, PD20B 1st Offspring ] x Adjusted for age, sex, height, weight, and smoking status. xx Standard error is corrected by GEE model.
Using intermediate phenotypes is an important tool for dissecting a complex disease into simpler traits and this strategy offers several advantages over using the disease itself. First, some intermediate phenotypes are quantitative in nature, and studies of such phenotypes could be more powerful if they fit the underlying genetics model better. Second, intermediate phenotypes are usually assessed using objective means, which can reduce the likelihood of misclassification. Third, comparing with the disease itself, intermediate phenotypes could involve fewer genes; hence, by dissecting a complex disease into more homogenous TABLE 5. Logistic GEE regression analysis of PD20B* of subsequent offspring in relation to their family members with interaction terms Adjustedx
Crude Model**
OR
95%CIxx
OR
95%CI
Paternal Father Mother 1st sib Father &1st sib
1.72 1.82 2.66 1.74
(0.72, 4.07) (0.94, 3.50) (1.21, 5.86) (0.84, 5.27)
1.76 2.46 2.08 1.92
(0.70, 4.40) (1.21, 5.00) (0.90, 5.85) (0.73, 5.06)
Maternal Father Mother 1st sib Mother &1st sib
1.19 1.84 2.02 3.22
(0.59, 2.24) (0.78, 4.36) (0.88, 4.64) (1.34, 7.78)
1.37 2.40 1.65 4.12
(0.66, 2.86) (0.92, 6.24) (0.67, 4.07) (1.72, 9.87)
Z is the matrix of environmental factors. X is the matrix of [PD20Bfather, PD20Bmother, PD20B1st Offspring, Interaction] *PD20B Z 1 indicates that the individual experienced a R 20% drop of FEV1 during the mefhacoline challenge test. **Model: LogitðPD20 BSubsequent Offspring Þ ¼ b1 XCb2 ZC3 x Adjusted for age, sex, height, weight, and smoking status. xx Standard error is corrected by GEE model.
FIGURE 2. Interaction of father/mother and the 1st offspring’s PD20B in predicting the outcome of subsequent offspring. Z-axis denotes the odds ratio of subsequent offspring being MTCH test positive. MTCHC denotes positive methacoline challenge test results (PD20B Z 1) and MTCH denotes negative methacoline challenge test results (PD20B Z 0). **Statistically significant at 0.05 level.
traits, we increase our chance of identifying the underlying genes. Previous studies showed a familial aggregation on asthma (18, 25). Our work focused on one intermediate phenotype underlying asthma, and has the advantages of improved statistical power and better control of potential confounding variables. In addition, our use of a randomly recruited, community-based family sample representing the general population reduces the likelihood of ascertainment bias, which could compromise a genetic epidemiological study. Many familial aggregation studies use a pair-wise correlation coefficient (20, 21). The random effect approach used in the current study has a number of desirable features
742
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
over traditional pair-wise design: 1) It avoids the arbitrary ordering of the sib pairs, where in pair-wise design ordering the offspring by age does not reflect their genetic relationship. The random effect model solves this problem elegantly by treating sibs in a symmetric manner; 2) The fixed effects of covariates can be simultaneously estimated with random effects entering in the covariance structure; 3) The covariance structure can be flexibly customized by the underlying assumption; 4) More than two offspring in each family can be easily accommodated into the model; 5) All correlation coefficients can be tested against the null simultaneously in one model. In the logistic regression analysis, we selected the AHR of younger offspring as the independent variable, assuming that all the offspring in each family were genetically equivalent. We chose this strategy because, 1) it simplifies the model and analysis and 2) it was clinically more meaningful to use parents and the older offspring’s phenotypes to predict the risk of AHR of the subsequent offspring. Classification of AHR could be confounded by the existence of certain environmental factors, such as airway infection, smoking, and the use of anti-asthmatic medication. During enrollment, we excluded subjects currently suffering from a cold or other airway infections, and we did not recruit subjects who were taking anti-asthma medicines. Current and former smokers were included in our study, and we controlled for their potential effects in the analysis. We also believe our measurement of AHR to methacoline test is more objective and reliable than asthma phenotypes assessed via self-reports or diagnoses by local doctors. The positive correlation coefficients found for all types of relative pairs and the stronger correlation with parentoffspring and offspring-offspring pairs than father-mother pairs suggest that genetic components may play a role. However, the extent of the genetic contribution is difficult to estimate accurately because familial clustering may be attributable to variation of multiple genes (4, 26, 27). AHR to methacoline is a well-studied biological pathway in the etiology of asthma. Contraction of respiratory smooth muscle is primarily controlled by the muscarinic acetylcholine receptor system, although it could be modified by other factors, such as immunologic mediators. Candidate gene studies have helped to explain some of the variance in AHR. For example, genes for a high-affinity receptor for immunoglobulin E (FcepsilonRI- beta) and b2-adrenoreceptor (B2AR) have been shown to be associated with AHR, (28, 29) and several other studies have linked AHR to a few chromosomal regions (30–32). However, most of the linkage signals did not reach the whole genome significance level and their results differ greatly across studies and across ethnic groups. This may reflect the multigenic nature of asthma and the relatively small contribution of each gene associated with the disease.
AEP Vol. 15, No. 10 November 2005: 737–743
Findings of this familial aggregation study support the need for additional investigations in searching for the genetic factors controlling AHR. Several chromosomal regions have been localized in previous linkage studies (30–32). A large number of genes have been identified in these suggestive linkage regions, and many of those that are functionally related to asthma and its intermediate phenotypes. These genes could serve as good candidate genes, and it is a promising area of research to test the association between functional genetic variations and asthma phenotypes. We gratefully acknowledge the assistance and cooperation of the faculty and staff of the Anhui Medical University, Anqing Public Health Bureau, and Anqing Hospital. We thank Dr. Scott Weiss for his insightful suggestions and comments on the current study.
REFERENCES 1. Laitinen T, Rasanen M, Kaprio J, Koskenvuo M, Laitinen LA. Importance of genetic factors in adolescent asthma: A population-based twin-family study. Am J Respir Crit Care Med. 1998;157:1073–1078. 2. Sheffer AL. Management of the adult asthma patient. Allergy Proc. 1995; 16:1–4. 3. Haahtela T, Lindholm H, Bjorksten F, Koskenvuo K, Laitinen LA. Prevalence of asthma in Finnish young men. BMJ. 1990;301:266–268. 4. Weeks DE, Lathrop GM. Polygenic disease: Methods for mapping complex disease traits. Trends Genet. 1995;11:513–519. 5. Burrows B, Sears MR, Flannery EM, Herbison GP, Holdaway MD. Relations of bronchial responsiveness to allergy skin test reactivity, lung function, respiratory symptoms, and diagnoses in thirteen-year-old New Zealand children. J Allergy Clin Immunol. 1995;95:548–556. 6. Marsh DG, Neely JD, Breazeale DR, Ghosh B, Freidhoff LR, EhrlichKautzky E, et al. Linkage analysis of IL4 and other chromosome 5q31.1 markers and total serum immunoglobulin E concentrations. Science. 1994;264:1152–1156. 7. Sandford AJ, Shirakawa T, Moffatt MF, Daniels SE, Ra C, Faux JA, et al. Localisation of atopy and beta subunit of high-affinity IgE receptor (Fc epsilon RI) on chromosome 11q. Lancet. 1993;341:332–334. 8. Burrows B, Martinez FD, Halonen M, Barbee RA, Cline MG. Association of asthma with serum IgE levels and skin-test reactivity to allergens. N Engl J Med. 1989;320:271–277. 9. Zimmerman B, Enander I, Zimmerman R, Ahlstedt S. Asthma in children less than 5 years of age: Eosinophils and serum levels of the eosinophil proteins ECP and EPX in relation to atopy and symptoms. Clin Exp Allergy. 1994;24:149–155. 10. Bousquet J, Chanez P, Vignola AM, Lacoste JY, Michel FB. Eosinophil inflammation in asthma. Am J Respir Crit Care Med. 1994;150:S33–38. 11. Ludviksdottir D, Janson C, Bjornsson E, Stalenheim G, Boman G, Hedenstrom H, et al. Different airway responsiveness profiles in atopic asthma, nonatopic asthma, and Sjogren’s syndrome. BHR Study Group. Bronchial hyperresponsiveness. Allergy. 2000;55:259–265. 12. Cockcroft DW, Killian DN, Mellon JJ, Hargreave FE. Bronchial reactivity to inhaled histamine: A method and clinical survey. Clin Allergy. 1977;7:235–243. 13. Hargreave FE, Ryan G, Thomson NC, O’Byrne PM, Latimer K, Juniper EF, et al. Bronchial responsiveness to histamine or methacholine in asthma: Measurement and clinical significance. J Allergy Clin Immunol. 1981; 68:347–355.
AEP Vol. 15, No. 10 November 2005: 737–743
14. O’Connor GT, Sparrow D, Weiss ST. Normal range of methacholine responsiveness in relation to prechallenge pulmonary function. The Normative Aging Study. Chest. 1994;105:661–666. 15. Rijcken B, Schouten JP, Weiss ST, Meinesz AF, de Vries K, van der Lende R. The distribution of bronchial responsiveness to histamine in symptomatic and in asymptomatic subjects. A population-based analysis of various indices of responsiveness. Am Rev Respir Dis. 1989;140:615–623. 16. Sparrow D, O’Connor GT, Rosner B, Weiss ST. Predictors of longitudinal change in methacholine airway responsiveness among middle-aged and older men: The Normative Aging Study. Am J Respir Crit Care Med. 1994;149:376–381. 17. Xu X, Niu T, Chen C, Wang B, Jin Y, Yang J, Weiss ST. Association of airway responsiveness with asthma and persistent wheeze in a Chinese population. Chest. 2001;119:691–700. 18. Jenkins MA, Hopper JL, Giles GG. Regressive logistic modeling of familial aggregation for asthma in 7,394 population-based nuclear families. Genet Epidemiol. 1997;14:317–332. 19. Palmer LJ, Burton PR, James AL, Musk AW, Cookson WO. Familial aggregation and heritability of asthma-associated quantitative traits in a population-based sample of nuclear families. Eur J Hum Genet. 2000; 8:853–860.
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
743
23. Jennrich R, Schluchter M. Unbalanced repeated-measures models with structured covariance matrices. Biometrics. 1986;42:805–820. 24. Wolfinger R, Tobias R, Sall J. Computing Gaussian likelihood and their derivatives for general linear mixed models. SIAM J Sci Comput. 1994;15:1294–1310. 25. von Mutius E, Nicolai T. Familial aggregation of asthma in a South Bavarian population. Am J Respir Crit Care Med. 1996;153:1266–1272. 26. Johnson GC, Todd JA. Strategies in complex disease mapping. Curr Opin Genet Dev. 2000;10:330–334. 27. Hampe J, Wienker T, Nurnberg P, Schreiber S. Mapping genes for polygenic disorders: Considerations for study design in the complex trait of inflammatory bowel disease. Hum Hered. 2000;50:91–101. 28. Ramsay CE, Hayden CM, Tiller KJ, Burton PR, Goldblatt J, Lesouef PN. Polymorphisms in the beta2-adrenoreceptor gene are associated with decreased airway responsiveness. Clin Exp Allergy. 1999;29:1195–1203. 29. Laprise C, Boulet LP, Morissette J, Winstall E, Raymond V. Evidence for association and linkage between atopy, airway hyper- responsiveness, and the beta subunit Glu237Gly variant of the high-affinity receptor for immunoglobulin E in the French-Canadian population. Immunogenetics. 2000;51:695–702.
20. Xu X, Yang J, Chen C, Wang B, Jin Y, Fang Z, et al. Familial aggregation of pulmonary function in a rural Chinese community. Am J Respir Crit Care Med. 1999;160:1928–1933.
30. Dizier MH, Besse-Schmittler C, Guilloud-Bataille M, Annesi-Maesano I, Boussaha M, Bousquet J, et al. Genome screen for asthma and related phenotypes in the French EGEA study. Am J Respir Crit Care Med. 2000;162:1812–1818.
21. Niu T, Rogus JJ, Chen C, Wang B, Yang J, Fang Z, et al. Familial aggregation of bronchodilator response: A community-based study. Am J Respir Crit Care Med. 2000;162:1833–1837.
31. Holroyd KJ, Martinati LC, Trabetti E, Scherpbier T, Eleff SM, Boner AL, et al. Asthma and bronchial hyperresponsiveness linked to the XY long arm pseudoautosomal region. Genomics. 1998;52:233–235.
22. Chatham M, Bleecker ER, Norman P, Smith PL, Mason P. A screening test for airways reactivity. An abbreviated methacholine inhalation challenge. Chest. 1982;82:15–18.
32. Daniels SE, Bhattacharrya S, James A, Leaves NI, Young A, Hill MR, et al. A genome-wide search for quantitative trait loci underlying asthma. Nature. 1996;383:247–250.
PURPOSE: We investigated the familial aggregation of airway hyper-responsiveness (AHR) to methacoline among randomly chosen families in a rural community in Anqing, China. METHODS: Airway responsiveness (AR) to methacoline and related risk factors were assessed in each subject. We first modeled the within family correlation in AR and demonstrated the familial aggregation of this trait. Furthermore, we examined the effect size (e.g., odds ratio, OR) of this correlation in a ‘‘subsequent offspring model.’’ RESULTS: The correlation coefficient is significantly positive for parent-offspring and offspring-offspring pairs, but not significant in father-mother pairs, suggesting a genetic component. The strength of the relationships is in the order of father-offspring ! mother-offspring ! offspring-offspring. The OR of a positive AHR test for subsequent offspring who had mothers and an eldest sibling with positive AHR is 4.12 (95% CI, 1.72–9.87), compared with subsequent offspring whose mother and eldest sibling were negative in the test. CONCLUSION: Our study supports a familial clustering of AHR in a Chinese population, which points to a role for genetic factors. Ann Epidemiol 2005;15:737–743. Ó 2005 Elsevier Inc. All rights reserved. KEY WORDS:
Familial Aggregation, Genetic Epidemiology, Asthma, Airway Responsiveness.
INTRODUCTION Many genetic and environmental factors could be involved in the development of asthma (1), a complex disease with the following characteristics: 1) airway obstruction that is reversible, either spontaneously or with treatment; 2) airway inflammation; and 3) airway hyperresponsiveness (AHR) to a variety of stimuli (2, 3). Like other complex disorders, asthma does not follow a simple Mendelian inheritance pattern (4). In mapping the asthma susceptibility gene, several intermediate phenotypes are used, including AHR (5, 6), total or specific immuno-globulin E (IgE) (7, 8), and blood eosinophil count (EOS) (9, 10). AHR, which has been considered a central feature of the definition of asthma, is usually regarded as an abnormal response of the lower respiratory tract to a number of nonsensitizing bronchoconstrictive stimuli (11–14). It is well recognized that there are significant inter-individual variations in airway responsiveness (AR) (15, 16) and AR is closely associated with asthma (17). Before focusing on the
From the Program for Population Genetics, Harvard School of Public Health, Boston, MA (K.H., C.C., B.W., X.X.); and Institute for Biomedical Research, Anhui Medical University, Hefei, China (J.Y., Z.F., X.X.). Address correspondence and reprint requests to: Xiping Xu, M.D., Ph.D., Program for Population Genetics, Harvard School of Public Health, 665 Huntington Ave. FXB-101, Boston, MA 02115. E-mail: xu@hsph. harvard.edu Received May 6, 2004; accepted February 1, 2005. Ó 2005 Elsevier Inc. All rights reserved. 360 Park Avenue South, New York, NY 10010
goal of locating susceptibility genes controlling asthma, a critical need to be addressed is, ‘‘does the disease cluster in families?’’ Using regressive logistic models, Jenkins et al. (18) showed positive familial aggregation of self-reported or clinic diagnosed asthma, and Palmer et al. (19) estimated the narrow sense heritability of lung function and doseresponsive slope to methacholine (MTCH) in the Busselton population, a population with a high asthma prevalence (20.2%). Xu et al. (20) found significant familial aggregation of lung function parameters (force vital capacity, FVC and forced expiratory volume in 1st second, FEV1) among asthma index and randomly chosen families. The familial aggregation of AHR, an important clinical characteristic of asthma, has not been well studied in a general population. To address this issue, we randomly selected 180 nuclear families from Anqing, China, and collected comprehensive phenotypic information. Compared with ascertained asthma families, these randomly selected families are more suitable for determining the familial inheritance patterns of quantitative intermediate phenotypes. We first modeled the within family correlation to examine the familial aggregation. Then we studied the relationship of AHR between subsequent offspring and their parents and elder sibling. This study helps clarify our understanding of the genetic epidemiology of AHR, and it may provide preliminary evidence for mapping genes underlying the intermediate phenotypes for asthma. 1047-2797/05/$–see front matter doi:10.1016/j.annepidem.2005.02.002
738
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
MATERIALS AND METHODS Study Subjects Anqing city is located in the central part of China with a population of 5.8 million in 1990. We chose Anqing as the study site for the following reasons: 1) the inhabitants are relatively homogeneous with respect to ethnicity, environment, occupation, and diet; 2) the stable population has existed for thousands of years; and 3) a uniform medical care service network has been established for more than 25 years, which offers a unique opportunity for efficient identification of study families. We selected families from the general population according to 1992 census records by a two-stage random sampling technique. At the first stage, we randomly selected villages from Anqing, and at the second stage, we randomly sampled one household in each selected village. Our inclusion criteria were: 1) a family size of at least four; 2) the availability of both parents; 3) two or more siblings in the family; and 4) at least 6 years old for the youngest offspring. Only the families that provided written informed consents were enrolled into the study. A detailed description of the study design and sample selection procedure has been presented elsewhere (20). Field Study Procedure Subject enrollment started in February 1995, after obtaining local China IRB approval. Brigham and Women’s Hospital IRB approval was received in September 1995. The field study was conducted by faculty members from Anhui Medical University, focusing on asthma and its intermediate phenotypes. The following procedures were carried out when the study subjects were invited to the research center: 1) standardized questionnaire (modified AST-DLD) assessing the respiratory history and symptoms, occupational and smoking histories, home environment, family history of asthma, and other chronic or genetic diseases; 2) pulmonary function test (spirometry); 3) methacholine challenge test for all subjects with FEV1 values O 60% of predicted value; and 4) bronchodilator test. In addition, weight and height were measured by standard methods after the subjects removed their shoes and outerwear. A detailed description of the study procedures has been presented elsewhere (20, 21). Measurement of Airway Hyper-responsiveness If baseline FEV1 was O 60% of predicted value, the airway challenge test was performed. A standard protocol for methacholine challenge test was developed based on a modified Chai protocol (22). Briefly, there were up to five steps to accomplish the methacholine challenge test with the following breath/dose schedules: saline solution, 5 mg/mL (one breath), 5 mg/mL (four breaths), 25 mg/mL (one breath), 25 mg/mL (four breaths). At each dose, two
AEP Vol. 15, No. 10 November 2005: 737–743
additional satisfactory spirometry maneuvers were obtained. The methacholine challenge test was continued until either a 20% drop of FEV1 from baseline was observed or the highest dose of the methacholine was reached, and then a following bronchodilator test was performed (22). We analyzed the dose–response relationship between methacholine dosage and FEV1 response of each subject in both continuous and binary forms. The continuous outcome is defined as the DFEV1/dose, where DFEV1 was defined as the drop FEV1 after the last methacholine challenge compared with the saline baseline, and dose was defined as the cumulative dosage methacholine administered. This variable was named as dose response slope (DRS) by convention. The binary outcome was defined as whether a subject showed a larger than or equal to 20% drop of FEV1 in the whole process, and was named PD20B. Statistical Analyses The families were randomly selected, and represented an unbiased and random sample of the general population. The DRS is strongly skewed to the right, and can be slightly negative, which means that FEV1 may increase after methacholine inhalation. This could be attributable to the biological phenomenon that the FEV1 bears certain degree of variation from time to time or from test to test. For example, the second measurement could give a slightly larger value than the first measurement, and resulted in a negative DRS. Another possibility is the learning effect that the participant performed better in the later maneuvers than the earlier ones. We somewhat controlled this confounder by letting the participants practice the spirometry a few times prior to the formal tests. A small number (1 unit) was added to all the DRS to make them all positive, and a log transformation is applied (Fig. 1). To test the impact of a few outliers, we conducted analysis by removing or including subjects outside the range defined by MeanDRS G 3SDDRS, and obtained similar results (data not shown). In this article we only present the analysis with all data points. Trait correlation among family members indicates the familial clustering. We modeled this correction using four parameters r1, r2, r3, and r4 standing for the father-mother, father-offspring, mother-offspring, and offspring-offspring correlation coefficients, respectively. In fitting the random effect model using a statistical software package (SAS, version 6.2), we specified an unstructured covariance matrix in order to estimate all these parameters. Y ¼ XbCZbC3; where Y denotes the outcome variable vector and X denotes the fixed effect variable matrix including all the potential
AEP Vol. 15, No. 10 November 2005: 737–743
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
739
FIGURE 1. The distribution of DRS and its log-transformation. Upper: The distribution of DRS. Lower: The distribution of Log(DRS). X-axis denotes the scale of the variable, and y-axis denotes the number of observation.
confounders including age, height, weight, gender, and smoking status. The random effect variable Z is a 775 ! 3 matrix (775 individuals in analysis). For any father, Zi has the value of [1,0,0]; for any mother, Zi has the value of [0,1,0]; and for any offspring, Zi has the value of [0,0,1]. The choice of Z specifies the structure of the within family correlation matrix. b and b are the fixed-effects and randomeffects parameters. The random effect model assumed b and 3 followed independent normal distributions.
E
b 0 ¼ 3 0
Var
b G0 ¼ 3 0R
predict the outcome of the subsequent offspring. The prediction value of family member’s AHR can also be considered as a measurement of the familial aggregation, and the odds ratio (OR) reflects the effect size. We fitted a ‘‘subsequent offspring model,’’ where the log(DRS) or PD20B of subsequent offspring were treated as dependent variable, and the value of his parents and the 1st offspring entered the model as covariates. A small number of families contributed two subsequent offspring, whose DRS or PD20B were hence not independent. Therefore, we applied a generalized estimation equation (GEE) with ‘‘exchangeable’’ correlation structure to adjust the standard error. RESULTS
The decomposition of V (variance of Y) was V Z ZGZ’ C R. We estimated G and R using restricted/ residual maximum likelihood (REML) (23), implemented in SAS PROC MIXED. The log-likelihood 1 1 lðG; RÞ ¼ log jVj log jX0 V1 Xj 2 2 1 0 1 np logð2pÞ; rV r 2 2 where r Z Y X(X#V1X)1X’V1Y and p is the rank of X, was maximized using the Newton-Raphson algorithm, and G and R were therefore estimated in the procedure (24). Upon detected significant familial aggregation, we were interested in using parents and/or the 1st offspring’s AHR to
The study included 180 families with complete data on pulmonary function and major covariates. The clinical characteristics of the study population are presented in Table 1. On average, fathers were 3.4 years older than mothers, and were more likely to be smokers than mothers and their offspring. The prevalence of clinically diagnosed asthma was low in this cohort (2.1 %). The average DRS and the percentage of positive methacoline test results (PD20B Z 1) were nearly the same among fathers, mothers, and children. First, we tested familial aggregation using a random effect model estimating the correlation coefficients for DRS among family members (Table 2). Fathers and mothers were slightly positively correlated but not significantly (cor Z 0.06), which might due to shared
740
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
AEP Vol. 15, No. 10 November 2005: 737–743
TABLE 1. Clinical characteristics of the 180 random families Mean (SD) or Number (percentage)
Father (n Z 180)
Mother (n Z 180)
First offspring (n Z 180)
Subsequent offspring (n Z 235)
44.4 (7.6) 1.65 (0.06) 57.4 (7.1) 0% 4.38 (0.65) 3.34 (0.60) 14 (7.7%) 142 (78.9%) 7 (3.9%) 1.01 (0.93) 48 (26.7%)
40.9 (7.8) 1.54 (0.05) 51.7 (6.7) 100% 3.37 (0.51) 2.66 (0.45) 0 3 (1.7%) 1 (0.5%) 0.93 (0.93) 44 (24.4%)
17.9 (7.0) 1.52 (0.13) 44.7 (12.2) 48.3% 3.39 (0.99) 2.90 (0.80) 2 (1.1%) 20 (11.1%) 4 (2.2%) 0.92 (0.95) 50 (27.7%)
15.6 (7.2) 1.44 (0.17) 38.5 (14.1) 47.2% 2.91 (1.09) 2.49 (0.88) 0 13 (5.5%) 4 (1.7%) 1.00 (1.02) 72 (30.6%)
Age, y Height, m Weight, kg Female, % Baseline FVC*, L Baseline FEV1**, L Former smoker Current smoker Diagnosed asthma, % Log (DRSx C1) PD20Bxx, %
*Forced vital capacity. **Forced expiratory volume in first second. x Methacoline dose–response slope. xx PD20B indicates that the individual experienced a R 20% drop of FEV1 during the methacoline challenge test.
environment factors. The correlation between fatheroffspring (cor Z 0.14, p ! 0.02) and mother-offspring (cor Z 0.14, p ! 0.02) pairs was comparable in our study. The highly significant positive correlation between siblings (cor Z 0.21, p ! 0.002) was observed. Second, a linear GEE regression model was fitted to test the relation of trait values between subsequent offspring and their parents and eldest available brother or sister. We found that the strength of the relation to subsequent offspring was in the same order as we observed in the random effect model: father ! mother ! eldest available offspring. The regression coefficient (b) of DRS between subsequent offspring and their eldest sibling was statistically significant in the models, although the magnitude of b was somewhat reduced in multivariate models due to the correlation of DRS among family members (Table 3). To diagnose or classify asthmatic status, the binary form of AHR is widely used and may be more meaningful in clinical practice. We applied logistic regression models and found the PD20B of subsequent offspring is significantly associated with that of their mother (OR Z 2.81; 95% CI, 1.39–5.70) and their eldest available siblings (OR Z 2.08; 95% CI, 1.06–4.07). The effect of the eldest available offspring was somewhat attenuated and becomes nonsignificant in the multivariate model (Table 4). We then TABLE 2. Correlation coefficients among family members (A family of two parents and three offspring) Correlation
Father
Mother
Offspring 1
Father Mother Offspring 1 Offspring 2 Offspring 3
1 0.063 0.138* 0.138* 0.138*
1 0.142* 0.142* 0.142*
1 0.206** 0.206**
*Statistically significant at 0.05 level. **Statistically significant at 0.01 level.
Offspring 2
1 0.206**
Offspring 3
1
introduced interaction terms into the model to further capture the familial aggregation pattern, and detected a significant interaction between the trait values of mother and those of her first available offspring (Table 5 and Fig. 2). The OR of having a positive AHR result for subsequent offspring who had a mother and elder sibling with positive AHR results was 4.12 (95% CI, 1.72–9.87). DISCUSSION In our study, correlation coefficients of family-relative pairs were significantly greater than zero, indicating a pattern of familial aggregation, which can be attributed to the sharing of genetic or environmental factors or both. The correlation was stronger in offspring-offspring and parent-offspring pairs than in parent-parent pairs, suggesting that genetic components may play a role in AHR. TABLE 3. Linear GEE regression analysis of log(DRS) of subsequent offspring in relation to their family members Adjustedx
Crude Model*
b
95% CI**
b
95%CI
Univariate Father Mother 1st sib
0.05 0.08 0.22
(0.08, 0.18) (0.02, 0.18) (0.09, 0.35)
0.08 0.15 0.20
(0.05, 0.22) (0.03, 0.26) (0.05, 0.34)
Multivariate Father Mother 1st sib
0.02 0.04 0.21
(0.10, 0.14) (0.04, 0.14) (0.07, 0.34)
0.05 0.10 0.14
(0.08, 0.18) (0.01, 0.21) (0.01, 0.27)
*Model: LogðDRSSubsequent Offspring Þ ¼ b1 XCb2 ZC3 Z is the matrix of environmental factors. In univariate model: X is the vector of [Log(DRSFather)i or Log(DRSmother)i or Log(DRS1st Offspring)i] In multivariate model: X is the matrix of [Log(DRSFather)i, Log(DRSmother)i, Log(DRS1st Offspring)i] **Standard error is corrected by GEE model. x Adjusted for age, sex, height, weight, and smoking status.
AEP Vol. 15, No. 10 November 2005: 737–743
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
741
TABLE 4. Logistic GEE regression analysis of PD20B* of subsequent offspring in relation to their family members Adjustedx
Crude Model**
OR
95%CIxx
OR
95%CI
Univariate Father Mother 1st sib
1.44 2.04 2.20
(0.74, 2.39) (1.08, 3.86) (1.16, 4.13)
1.62 2.81 2.08
(0.81, 2.73) (1.39, 5.70) (1.06, 4.07)
Multivariate Father Mother 1st sib
1.19 1.75 1.92
(0.59, 2.38) (0.90, 3.38) (0.98, 1.92)
1.37 2.44 1.68
(0.67, 2.83) (1.19, 4.99) (0.83, 3.40)
*PD20B Z 1 indicates that the individual experienced a R 20% drop of FEV1 during the methacoline challenge test. **Model: LogitðPD20 BSubsequent Offspring Þ ¼ b1 XCb2 ZC3 Z is the matrix of environmental factors. In univariate model: X is [PD20B Father or PD20B mother or PD20B 1st Offspring] In multivariate model: X is the matrix of [PD20B Father, PD20B mother, PD20B 1st Offspring ] x Adjusted for age, sex, height, weight, and smoking status. xx Standard error is corrected by GEE model.
Using intermediate phenotypes is an important tool for dissecting a complex disease into simpler traits and this strategy offers several advantages over using the disease itself. First, some intermediate phenotypes are quantitative in nature, and studies of such phenotypes could be more powerful if they fit the underlying genetics model better. Second, intermediate phenotypes are usually assessed using objective means, which can reduce the likelihood of misclassification. Third, comparing with the disease itself, intermediate phenotypes could involve fewer genes; hence, by dissecting a complex disease into more homogenous TABLE 5. Logistic GEE regression analysis of PD20B* of subsequent offspring in relation to their family members with interaction terms Adjustedx
Crude Model**
OR
95%CIxx
OR
95%CI
Paternal Father Mother 1st sib Father &1st sib
1.72 1.82 2.66 1.74
(0.72, 4.07) (0.94, 3.50) (1.21, 5.86) (0.84, 5.27)
1.76 2.46 2.08 1.92
(0.70, 4.40) (1.21, 5.00) (0.90, 5.85) (0.73, 5.06)
Maternal Father Mother 1st sib Mother &1st sib
1.19 1.84 2.02 3.22
(0.59, 2.24) (0.78, 4.36) (0.88, 4.64) (1.34, 7.78)
1.37 2.40 1.65 4.12
(0.66, 2.86) (0.92, 6.24) (0.67, 4.07) (1.72, 9.87)
Z is the matrix of environmental factors. X is the matrix of [PD20Bfather, PD20Bmother, PD20B1st Offspring, Interaction] *PD20B Z 1 indicates that the individual experienced a R 20% drop of FEV1 during the mefhacoline challenge test. **Model: LogitðPD20 BSubsequent Offspring Þ ¼ b1 XCb2 ZC3 x Adjusted for age, sex, height, weight, and smoking status. xx Standard error is corrected by GEE model.
FIGURE 2. Interaction of father/mother and the 1st offspring’s PD20B in predicting the outcome of subsequent offspring. Z-axis denotes the odds ratio of subsequent offspring being MTCH test positive. MTCHC denotes positive methacoline challenge test results (PD20B Z 1) and MTCH denotes negative methacoline challenge test results (PD20B Z 0). **Statistically significant at 0.05 level.
traits, we increase our chance of identifying the underlying genes. Previous studies showed a familial aggregation on asthma (18, 25). Our work focused on one intermediate phenotype underlying asthma, and has the advantages of improved statistical power and better control of potential confounding variables. In addition, our use of a randomly recruited, community-based family sample representing the general population reduces the likelihood of ascertainment bias, which could compromise a genetic epidemiological study. Many familial aggregation studies use a pair-wise correlation coefficient (20, 21). The random effect approach used in the current study has a number of desirable features
742
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
over traditional pair-wise design: 1) It avoids the arbitrary ordering of the sib pairs, where in pair-wise design ordering the offspring by age does not reflect their genetic relationship. The random effect model solves this problem elegantly by treating sibs in a symmetric manner; 2) The fixed effects of covariates can be simultaneously estimated with random effects entering in the covariance structure; 3) The covariance structure can be flexibly customized by the underlying assumption; 4) More than two offspring in each family can be easily accommodated into the model; 5) All correlation coefficients can be tested against the null simultaneously in one model. In the logistic regression analysis, we selected the AHR of younger offspring as the independent variable, assuming that all the offspring in each family were genetically equivalent. We chose this strategy because, 1) it simplifies the model and analysis and 2) it was clinically more meaningful to use parents and the older offspring’s phenotypes to predict the risk of AHR of the subsequent offspring. Classification of AHR could be confounded by the existence of certain environmental factors, such as airway infection, smoking, and the use of anti-asthmatic medication. During enrollment, we excluded subjects currently suffering from a cold or other airway infections, and we did not recruit subjects who were taking anti-asthma medicines. Current and former smokers were included in our study, and we controlled for their potential effects in the analysis. We also believe our measurement of AHR to methacoline test is more objective and reliable than asthma phenotypes assessed via self-reports or diagnoses by local doctors. The positive correlation coefficients found for all types of relative pairs and the stronger correlation with parentoffspring and offspring-offspring pairs than father-mother pairs suggest that genetic components may play a role. However, the extent of the genetic contribution is difficult to estimate accurately because familial clustering may be attributable to variation of multiple genes (4, 26, 27). AHR to methacoline is a well-studied biological pathway in the etiology of asthma. Contraction of respiratory smooth muscle is primarily controlled by the muscarinic acetylcholine receptor system, although it could be modified by other factors, such as immunologic mediators. Candidate gene studies have helped to explain some of the variance in AHR. For example, genes for a high-affinity receptor for immunoglobulin E (FcepsilonRI- beta) and b2-adrenoreceptor (B2AR) have been shown to be associated with AHR, (28, 29) and several other studies have linked AHR to a few chromosomal regions (30–32). However, most of the linkage signals did not reach the whole genome significance level and their results differ greatly across studies and across ethnic groups. This may reflect the multigenic nature of asthma and the relatively small contribution of each gene associated with the disease.
AEP Vol. 15, No. 10 November 2005: 737–743
Findings of this familial aggregation study support the need for additional investigations in searching for the genetic factors controlling AHR. Several chromosomal regions have been localized in previous linkage studies (30–32). A large number of genes have been identified in these suggestive linkage regions, and many of those that are functionally related to asthma and its intermediate phenotypes. These genes could serve as good candidate genes, and it is a promising area of research to test the association between functional genetic variations and asthma phenotypes. We gratefully acknowledge the assistance and cooperation of the faculty and staff of the Anhui Medical University, Anqing Public Health Bureau, and Anqing Hospital. We thank Dr. Scott Weiss for his insightful suggestions and comments on the current study.
REFERENCES 1. Laitinen T, Rasanen M, Kaprio J, Koskenvuo M, Laitinen LA. Importance of genetic factors in adolescent asthma: A population-based twin-family study. Am J Respir Crit Care Med. 1998;157:1073–1078. 2. Sheffer AL. Management of the adult asthma patient. Allergy Proc. 1995; 16:1–4. 3. Haahtela T, Lindholm H, Bjorksten F, Koskenvuo K, Laitinen LA. Prevalence of asthma in Finnish young men. BMJ. 1990;301:266–268. 4. Weeks DE, Lathrop GM. Polygenic disease: Methods for mapping complex disease traits. Trends Genet. 1995;11:513–519. 5. Burrows B, Sears MR, Flannery EM, Herbison GP, Holdaway MD. Relations of bronchial responsiveness to allergy skin test reactivity, lung function, respiratory symptoms, and diagnoses in thirteen-year-old New Zealand children. J Allergy Clin Immunol. 1995;95:548–556. 6. Marsh DG, Neely JD, Breazeale DR, Ghosh B, Freidhoff LR, EhrlichKautzky E, et al. Linkage analysis of IL4 and other chromosome 5q31.1 markers and total serum immunoglobulin E concentrations. Science. 1994;264:1152–1156. 7. Sandford AJ, Shirakawa T, Moffatt MF, Daniels SE, Ra C, Faux JA, et al. Localisation of atopy and beta subunit of high-affinity IgE receptor (Fc epsilon RI) on chromosome 11q. Lancet. 1993;341:332–334. 8. Burrows B, Martinez FD, Halonen M, Barbee RA, Cline MG. Association of asthma with serum IgE levels and skin-test reactivity to allergens. N Engl J Med. 1989;320:271–277. 9. Zimmerman B, Enander I, Zimmerman R, Ahlstedt S. Asthma in children less than 5 years of age: Eosinophils and serum levels of the eosinophil proteins ECP and EPX in relation to atopy and symptoms. Clin Exp Allergy. 1994;24:149–155. 10. Bousquet J, Chanez P, Vignola AM, Lacoste JY, Michel FB. Eosinophil inflammation in asthma. Am J Respir Crit Care Med. 1994;150:S33–38. 11. Ludviksdottir D, Janson C, Bjornsson E, Stalenheim G, Boman G, Hedenstrom H, et al. Different airway responsiveness profiles in atopic asthma, nonatopic asthma, and Sjogren’s syndrome. BHR Study Group. Bronchial hyperresponsiveness. Allergy. 2000;55:259–265. 12. Cockcroft DW, Killian DN, Mellon JJ, Hargreave FE. Bronchial reactivity to inhaled histamine: A method and clinical survey. Clin Allergy. 1977;7:235–243. 13. Hargreave FE, Ryan G, Thomson NC, O’Byrne PM, Latimer K, Juniper EF, et al. Bronchial responsiveness to histamine or methacholine in asthma: Measurement and clinical significance. J Allergy Clin Immunol. 1981; 68:347–355.
AEP Vol. 15, No. 10 November 2005: 737–743
14. O’Connor GT, Sparrow D, Weiss ST. Normal range of methacholine responsiveness in relation to prechallenge pulmonary function. The Normative Aging Study. Chest. 1994;105:661–666. 15. Rijcken B, Schouten JP, Weiss ST, Meinesz AF, de Vries K, van der Lende R. The distribution of bronchial responsiveness to histamine in symptomatic and in asymptomatic subjects. A population-based analysis of various indices of responsiveness. Am Rev Respir Dis. 1989;140:615–623. 16. Sparrow D, O’Connor GT, Rosner B, Weiss ST. Predictors of longitudinal change in methacholine airway responsiveness among middle-aged and older men: The Normative Aging Study. Am J Respir Crit Care Med. 1994;149:376–381. 17. Xu X, Niu T, Chen C, Wang B, Jin Y, Yang J, Weiss ST. Association of airway responsiveness with asthma and persistent wheeze in a Chinese population. Chest. 2001;119:691–700. 18. Jenkins MA, Hopper JL, Giles GG. Regressive logistic modeling of familial aggregation for asthma in 7,394 population-based nuclear families. Genet Epidemiol. 1997;14:317–332. 19. Palmer LJ, Burton PR, James AL, Musk AW, Cookson WO. Familial aggregation and heritability of asthma-associated quantitative traits in a population-based sample of nuclear families. Eur J Hum Genet. 2000; 8:853–860.
Hao et al. FAMILIAL AGGREGATION OF AIRWAY RESPONSIVENESS
743
23. Jennrich R, Schluchter M. Unbalanced repeated-measures models with structured covariance matrices. Biometrics. 1986;42:805–820. 24. Wolfinger R, Tobias R, Sall J. Computing Gaussian likelihood and their derivatives for general linear mixed models. SIAM J Sci Comput. 1994;15:1294–1310. 25. von Mutius E, Nicolai T. Familial aggregation of asthma in a South Bavarian population. Am J Respir Crit Care Med. 1996;153:1266–1272. 26. Johnson GC, Todd JA. Strategies in complex disease mapping. Curr Opin Genet Dev. 2000;10:330–334. 27. Hampe J, Wienker T, Nurnberg P, Schreiber S. Mapping genes for polygenic disorders: Considerations for study design in the complex trait of inflammatory bowel disease. Hum Hered. 2000;50:91–101. 28. Ramsay CE, Hayden CM, Tiller KJ, Burton PR, Goldblatt J, Lesouef PN. Polymorphisms in the beta2-adrenoreceptor gene are associated with decreased airway responsiveness. Clin Exp Allergy. 1999;29:1195–1203. 29. Laprise C, Boulet LP, Morissette J, Winstall E, Raymond V. Evidence for association and linkage between atopy, airway hyper- responsiveness, and the beta subunit Glu237Gly variant of the high-affinity receptor for immunoglobulin E in the French-Canadian population. Immunogenetics. 2000;51:695–702.
20. Xu X, Yang J, Chen C, Wang B, Jin Y, Fang Z, et al. Familial aggregation of pulmonary function in a rural Chinese community. Am J Respir Crit Care Med. 1999;160:1928–1933.
30. Dizier MH, Besse-Schmittler C, Guilloud-Bataille M, Annesi-Maesano I, Boussaha M, Bousquet J, et al. Genome screen for asthma and related phenotypes in the French EGEA study. Am J Respir Crit Care Med. 2000;162:1812–1818.
21. Niu T, Rogus JJ, Chen C, Wang B, Yang J, Fang Z, et al. Familial aggregation of bronchodilator response: A community-based study. Am J Respir Crit Care Med. 2000;162:1833–1837.
31. Holroyd KJ, Martinati LC, Trabetti E, Scherpbier T, Eleff SM, Boner AL, et al. Asthma and bronchial hyperresponsiveness linked to the XY long arm pseudoautosomal region. Genomics. 1998;52:233–235.
22. Chatham M, Bleecker ER, Norman P, Smith PL, Mason P. A screening test for airways reactivity. An abbreviated methacholine inhalation challenge. Chest. 1982;82:15–18.
32. Daniels SE, Bhattacharrya S, James A, Leaves NI, Young A, Hill MR, et al. A genome-wide search for quantitative trait loci underlying asthma. Nature. 1996;383:247–250.