Statistical/Design Methods

ERIKA VON MUTIUS

University Children's Hospital Munich, Munich, Germany

	INTRODUCTION

TOP INTRODUCTION WHAT DO WE NEED... HOW DO WE ACHIEVE... REFERENCES

Numerous studies have shown that asthma is associated with inflammatory changes in the airways, and various techniques have been proposed to measure airway inflammation. Besides the problems encountered with the various measurement techniques and the accessibility of the airways, particularly in young children, some peculiarities of the natural course of childhood wheezing illnesses must be taken into account when designing studies of airway inflammation in young children.

First, there is increasing evidence that asthma starts early in life. A large U.S. survey conducted in Rochester, Minnesota, tracking back medical records into early childhood, demonstrated that in most adults with asthma the disease started in the first 1 to 4 yr of life (1). Moreover, birth cohort studies in the United States and Europe have documented that children with persistent wheeze at school age and characteristics of asthma such as atopy and airway hyperresponsiveness start wheezing in their first years of life (2, 3). However, not all that wheezes develops asthma. Benign phenotypes of infant wheezing illnesses have been identified in which children outgrow their symptoms around their third birthday (3). These wheezing phenotypes are associated with different characteristics such as maternal smoking in pregnancy and postnatally diminished lung function in the infant. Finally, a third category of wheezing illness may exist that occurs mostly with viral infections, is not related to atopy and airway hyperresponsiveness, and has a better prognosis during school-age and adolescent years than atopy-related asthma (4, 5).

Experimental and epidemiological studies (6) also indicate that the development of atopic childhood asthma and the inception of processes closely related to asthma such as atopic sensitization are determined early on in life (9). Peat and colleagues (9) have demonstrated that the pattern of atopic sensitization associated with childhood asthma is characterized by early occurrence, probably developing much earlier than the eighth year of life. Such findings are in agreement with the results of the Tucson birth cohort study showing a significant association between increased serum immunoglobulin (Ig) E levels at the age of 9 mo, but not at birth, and the subsequent development of asthma at school age (3). These results suggest that environmental factors occurring in the first months of life and/or in utero regulate total serum IgE levels, which in turn are closely related to the manifestation of childhood asthma (10). Studies have also shown that immunological recognition of environmental allergens begins before birth (6, 7), indicating that factors affecting prenatal exposure and the subsequent maternofetal immune response in utero may also play an important role in the subsequent development of childhood asthma. Thus, the window of opportunity for environmental influences to affect the development of childhood asthma may be restricted to antenatal or infant exposure.

	WHAT DO WE NEED TO KNOW?

TOP INTRODUCTION WHAT DO WE NEED... HOW DO WE ACHIEVE... REFERENCES

Airway inflammation may result from noxious fumes, environmental pollutants, viral and bacterial infections, and allergen exposure in atopic subjects. A variety of inflammatory changes may be found in the airways, but there is evidence to suggest that certain specific exposures elicit characteristic inflammatory responses. Ozone is an example (11). Therefore, the association between various biomarkers and different wheezing illnesses and other respiratory conditions in early childhood should be studied to identify the peculiarities associated with certain conditions or exposures. Care should be taken to disentangle the various wheezing phenotypes and to consider the heterogeneity of asthma because both the pathogenesis and etiology of these conditions may differ substantially. Furthermore, it seems essential to investigate how markers of airway inflammation relate to the developmental aspects of a growing lung and a maturing immune system in infancy and early childhood because significant changes over time may occur.

	HOW DO WE ACHIEVE THIS?

TOP INTRODUCTION WHAT DO WE NEED... HOW DO WE ACHIEVE... REFERENCES

As there is a spectrum of various direct and indirect measures of airway inflammation, there is also a spectrum of epidemiological study designs. A substantial component of what has been called molecular epidemiology (12) involves studies that evaluate biomarkers to determine their optimal use in subsequent etiological studies. Such preliminary studies assess the reliability and, if at all possible, the accuracy of the assay(s) performed on the specimens, and optimize conditions for collecting, processing, and storing biological specimens before measurement. Of primary importance is assessment of the reproducibility of any laboratory assay.

Further studies will evaluate the distribution and the endogenous and exogenous determinants of those biomarkers in specified populations. These findings will determine how to apply the marker in subsequent etiological investigations. Cross-sectional and longitudinal characterization studies may be performed. Cross-sectional studies are useful for the evaluation of demographic variables such as age for independent effects on the levels of certain biomarkers. If biomarkers are known or suspected to vary over the duration of a given study period then longitudinal studies might be considered first before using such a marker in further etiological investigations.

The next step in evaluating the usefulness of a biological marker is to perform observational, cross-sectional, case-control, or longitudinal studies to assess the etiological relationship between a certain marker and the disease of interest. A case-control study includes subjects with the disease and a suitable control group. The relationship between a biological marker and the disease is examined by comparing how frequently the attribute is present, or how the levels differ if the marker is quantitative, in subjects with and without the disease. In the case of early childhood wheezing illnesses, there is currently no diagnostic tool to identify the various phenotypes. This implies that in cross-sectional studies enrolling young children with wheeze a variety of phenotypes will be present, obscuring potential associations with diagnostic markers, treatment outcomes, environmental risk factors, and genetic determinants. Information about the nature of the wheezing illness may be provided by careful characterization of study subjects with respect to their family history of asthma, hay fever, and atopic eczema; smoking status of the mother and other caregivers; presence of symptoms of atopy such as eczema; increased total and specific IgE levels; and the role of viral infections and other trigger factors. However, an unequivocal categorization of affected subjects cannot always be achieved.

Longitudinal studies are the most powerful designs, yet they are also the most complex, time consuming, and costly. A study population that is representative of the general population will maximize the generalizability of the study findings. In turn, a study population at high risk of developing the disease may be more efficient, but may also produce results that are less generalizable. With regard to early childhood wheezing illnesses, longitudinal approaches may prove particularly useful because case definition may be correctly achieved only over time, despite the presence of significant risk factors at birth or even before birth. To increase the efficiency of prospective studies, nested case-control designs might be used. These studies involve collecting information about all members of a cohort, storing data and biologic specimens, monitoring the cohort until the development or identification of the disease of interest, and then analyzing the stored material from the cases and the control subjects.

The aims of the studies must be declared beforehand and may fall into one of several categories: (1) the investigation of the natural history of pulmonary conditions associated with inflammatory changes in the airways, (2) studies of the mechanisms of disease, and (3) interventions, for example, randomized clinical trials addressing markers of inflammation as primary outcomes. In any case, sample size calculations based on a given combination of significance level, power, and size of expected effects should be performed before the start of the study. Appropriate statistical techniques accounting for small numbers, such as the Fisher exact test, should be used for comparisons between groups. In clinical applications, when using biomarkers to evaluate the underlying disease process, calculation of sensitivity, specificity, and predictive values and graphical illustration as receiver-operator curves may prove useful tools. Sensitivity may then be defined as the probability that a person with the condition will be classified correctly as having the condition on the basis of the marker, and specificity may be defined as the probability that a person without the marker will be classified correctly as not having the disease. All data must be examined further for the validity of the assumptions underlying the statistical techniques used.

When attempting to interpret the results of such studies several particularities of data analysis must be borne in mind. For example, odds ratios may cause semantic confusion. Odds ratios merely reflect the strength, but not the direction, of an association, although they are often designated as "risks." An increased risk of asthma associated with certain determinants may indicate a causal link whereby these factors induce the development of new cases of asthma. However, the opposite is equally conceivable, whereby the propensity to develop asthma also confers a susceptibility to develop certain characteristics such as, for example, atopic sensitization to house dust mites, which in itself is not a causal factor for the inception of asthma. Likewise, associations between certain inflammatory markers, in, for example, asthma or other pulmonary illnesses, may not reflect differences in the etiology of the condition, but may result from differences in the pathogenesis.

To date most studies have operated on a group level, where mean values for certain attributes are compared between groups and then generalized in an attempt to fit all individuals into respective groups. However, the prognosis of an individual patient can be expressed only in probability terms. With a better understanding of the genetic background of a subject, the individual susceptibility to certain harmful or beneficial exposures such as air pollution (11) or drugs (13) may be foreseen, thereby substantially improving our understanding of the effect of environmental exposure at an individual level and to design intervention strategies for particular patients.

	Footnotes

Correspondence and requests for reprints should be addressed to Erika von Mutius, M.D., University Children's Hospital, Lindwurmstrasse 4, Munich D080337, Germany. E-mail: erika.von.mutius{at}kk-imed.uni-muenchen.de

	References

TOP INTRODUCTION WHAT DO WE NEED... HOW DO WE ACHIEVE... REFERENCES

1. Yunginger, J. W., C. E. Reed, E. J. O'Connell, L. J. Melton, W. M. O'Fallon, and M. D. Silverstein. 1992. A community-based study of the epidemiology of asthma: incidence rates, 1964-1983. Am. Rev. Respir. Dis. 146: 888-894 [Medline].

2. Martinez, F. D., A. L. Wright, L. M. Taussig, C. J. Holberg, M. Halonen, W. J. Morgan, and for the Group Health Medical Associates. 1995. Asthma and wheezing in the first six years of life. N. Engl. J. Med. 332: 133-138 [Abstract/Free Full Text].

3. Sporik, R., S. T. Holgate, and J. J. Cogswell. 1991. Natural history of asthma in childhooda birth cohort study. Arch. Dis. Child. 66: 1050-1053 [Abstract].

4. Helms, P. J.. 1994. Editorial: wheezing infants. Clin. Exp. Allergy 2: 97-99 .

5. von Mutius, E., S. Illi, T. Hirsch, W. Leupold, U. Keil, and S. Weiland. 1999. Frequency of infections in the first years of life and risk of asthma, atopy and airway hyperresponsiveness among schoolage children. Eur. Respir. J. 14: 4-11 [Abstract].

6. Piccinni, M. P., F. Mecacci, S. Sampognaro, R. Manetti, P. Parronchi, E. Maggi, and S. Romagnani. 1993. Aeroallergen sensitisation can occur during fetal life. Int. Arch. Allergy Immunol. 102: 301-303 [Medline].

7. Warner, J. A., A. C. Jones, E. A. Miles, B. M. Colwell, and J. O. Warner. 1996. Maternofetal interaction and allergy. Allergy 51: 447-451 [Medline].

8. von Mutius, E., S. K. Weiland, C. Fritzsch, H. Duhme, and U. Keil. 1998. Increasing prevalence of hay fever and atopy among children in Leipzig, East Germany. Lancet 351: 862-866 [Medline].

9. Peat, J. K., C. M. Salome, and A. J. Woolcock. 1990. Longitudinal changes in atopy during a 4-year period: relation to bronchial hyperresponsiveness and respiratory symptoms in a population sample of Australian schoolchildren. J. Allergy Clin. Immunol. 85: 65-74 [Medline].

10. Sears, M. R., B. Burrows, E. M. Flannery, G. P. Herbison, C. J. Hewitt, and M. D. Holdaway. 1991. Relation between airway responsiveness and serum IgE in children with asthma and in apparently normal children. N. Engl. J. Med. 325: 1067-1071 [Abstract].

11. von Mutius, E. 1999. Air pollution and asthma. In J. Bousquet and H. Yssel, editors. Lung Biology in Health and Disease: Immunotherapy in Asthma. Marcel Dekker, New York. 497-534.

12. Schulte, P. A., and F. P. Perera, editors. 1993. Molecular Epidemiology. Principles and Practices. Academic Press, San Diego, CA.

13. Martinez, F. D., P. E. Graves, M. Baldini, S. Solomon, and R. Erickson. 1997. Association between genetic polymorphisms of the beta2-adrenoceptor and response to albuterol in children with and without a history of wheezing. J. Clin. Invest. 100: 3184-3188 [Medline].