INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Asthma is the most common chronic disease of childhood (1), and its prevalence and possibly its severity are increasing (2). This changing pattern of disease has not been explained, in part because of our incomplete understanding of the pathogenesis of the syndrome. Many risk factors have been identified (3), but it is not clear how these factors act to produce disease in any given individual. Although an atopic predisposition is one of the strongest risk factors for the development of asthma (4), many persons with atopy do not have lower respiratory symptoms. Other factors, in addition to atopy, must be operative before the asthma phenotype becomes manifest.

Respiratory infections have been clearly shown to precipitate acute attacks in persons with established asthma (5). However, the role of infection in the initiation of the syndrome is controversial (6). The universal exposure of the immature immune system to viral pathogens results in a high incidence of respiratory infection in the first year of life (7). Viral respiratory infections are frequently accompanied by wheezing in infants (7), and early reports have suggested that these wheezing episodes are a manifestation of asthma (8). However, Martinez and colleagues (9) have recently shown that only a minority of infants with wheezing in infancy has a physician diagnosis of asthma by age six. Furthermore, these investigators (10) and others (11) have suggested that viral infection could actually protect against the development of asthma.

One aspect of the problem, which has received little attention, is the role of a predisposing history of asthma/atopy as a risk factor for respiratory infection. Although prospective studies have documented an increased rate of viral isolations in asthmatics as compared with nonasthmatics (5, 12), similar studies in persons with atopy have failed to show an increased frequency of infection in those affected (13). Infants at risk of atopy/asthma, but with no manifest disease, have not been studied with surveillance methods. An opportunity to address some of these issues is provided by a study of nitrogen dioxide and respiratory illness in children in Albuquerque, New Mexico (16). In this investigation, infants were followed from birth until they were 18 mo of age with intensive surveillance of respiratory illnesses. The current analysis was undertaken to evaluate the relationship between the predisposition to atopy (measured as a parental history of allergy or asthma) and the development of various respiratory syndromes. We hypothesized that a parental history of allergy/asthma would be associated with an increased rate of respiratory infection. In addition, we hypothesized that infants with a parental history of allergy/ asthma would have an increased incidence of wheezing illness after adjusting for factors that increase the risk of infection.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Details of the study design have been published previously (16). In brief, the objective of the study was to determine the effects of nitrogen dioxide on the incidence and severity of respiratory illnesses in the first 18 mo of life. Nitrogen dioxide is an oxidant gas that results from high-temperature combustion and is increased in homes that contain gas cooking stoves. The incidence of respiratory symptoms and illness were not found to be related to 2-wk average concentrations of nitrogen dioxide levels or to the presence of gas stoves. These negative findings allow the combination of surveillance data from children from homes with gas and electric cooking stoves in the current analysis.

Infants were identified at birth from all full-term normal births (weight > 2,500 g) in Albuquerque, New Mexico between January 1, 1988 and June 30, 1990. Selection criteria included (1) an English-speaking mother at least 18 yr of age, (2) no household members who smoked, and (3) no plans for full-time day care (> 20 h/wk). A daily symptom diary was developed for completion by the infant's mother. Symptoms and signs included runny or stuffy nose, fever, dry cough, wet cough, wheezing, trouble breathing, trouble feeding, and rash; the mother also recorded if she considered the child to be "sick." At enrollment, the mothers were instructed in the use of the diary by either of two trained nurses. During the subsequent 18 mo of surveillance, the mother was telephoned every 2 wk by an interviewer. In standardized questions, the mother was asked to review the diary and to report all symptoms and any illness since the last call. If respiratory symptoms were reported for the child, information was obtained on physician visits, and outpatient records were later abstracted by one nurse using a standardized protocol. A nurse practitioner or nurse made home visits to evaluate symptomatic and randomly selected asymptomatic children. At the time of the follow-up telephone call, information about the status of breast feeding and attendance at day care was obtained. In addition, the mother was asked if she or anyone else in the family had been ill during the 2 wk prior to the call.

Illness events were defined as the occurrence of at least two consecutive days of runny or stuffy nose, wet cough, dry cough, wheeze, or trouble breathing. The illness events ended with two consecutive symptom-free days. The illness events were further classified as "upper" or "lower" respiratory tract. Upper respiratory tract illnesses (URI) included any combination of runny or stuffy nose, dry cough, or trouble breathing, and lower respiratory tract illnesses (LRI) had at least 1 d of reported wet cough or wheeze in addition to the symptoms of a URI. The LRIs were further stratified as wheezing LRI, if wheezing was reported for at least 1 d or as wet cough LRI if wheezing was absent.

In a previous publication (17) the diagnostic accuracy of the surveillance system was compared with the nurse's diagnosis and to the diagnostic labels given by physicians during clinical visits. Comparison of the parent's illness report with the nurse practitioner's diagnoses showed that the surveillance system had a sensitivity of 93.4% for lower respiratory illnesses, but a specificity of only 24%. Most of the false positive diagnoses of lower respiratory illness were associated with a wet cough; the specificity of lower respiratory illnesses with wheezing was much higher. Seventy-three percent of the infants whose parents described wheezing were classified as having a lower respiratory illness by the nurse practitioners. For this reason, the current data were analyzed separately for those infants with wheezing LRIs and wheezing and those with wet cough LRIs.

Information about risk factors was obtained at the beginning of follow-up using a standardized questionnaire. Information was obtained on the race and ethnicity of the parents, household income, numbers of hours the infant spent in day care, the number of older siblings, the living space in the home, and household composition. Both parents completed the respiratory symptoms questionnaire developed by the American Thoracic Society (18). Parents were categorized as having asthma if they reported a doctor's diagnosis for asthma, and they were considered atopic if they reported a history of hay fever or desensitization shots. A parental history of atopy was classified as positive if either or both parents were atopic and both denied asthma. A parental history of asthma was classified as positive if either parent had asthma regardless of atopic status.

Data Analysis

In calculating incidence rates of respiratory illness, days at risk were those days of observation during which an illness was not in progress. Incidence rates of the several types of illness were calculated as the ratio of the number of illness events to the number of days at risk, annualized to 365 days at risk. Confidence limits for the rates were calculated using an estimate of the standard error based on the Poisson distribution (19). Incidence rates were calculated for the entire group and for the subgroups defined by parental history: (1) no parental history of atopy/asthma, (2) a parental history of atopy but not asthma, (3) parental history of asthma. Tests of significance for illness rates by age group and family history of asthma or atopy were computed from Poisson regression models.

Multivariate logistic regression methods were used to model the occurrence of respiratory illness during each 2-wk interval of follow-up that the infant was at risk. Infant respiratory events were related to host factors previously shown to be associated with wheezing (age, sex, ethnicity, parental history of asthma/atopy) and environmental factors known to promote the transmission of infections (day care attendance, presence of older siblings, month of illness, crowding index, and contact with a sick family member). Contact with a sick family member was classified as positive if it occurred during the 2 wk preceding the infant's illness. The crowding index was calculated as square feet of home/person living in the home. Household income was used as an index of socioeconomic status, and breast-feeding practices were included in all of the models because of previous associations of these covariates with wheezing events and/or rates of respiratory infections (16).

Separate analyses were performed for the occurrence of any respiratory illnesses, any LRI, and LRI with wheezing. By our definitions, infants with LRI also had symptoms of URI. Therefore, the rate of all respiratory illnesses was used as an index of the total burden of respiratory infection in the infants. Because of the importance of the first 6 mo of life in the development of the immune system, the models were repeated for infants younger than 6 mo of age and for those 7 to 18 mo of age.

In order to account for the correlation between multiple measures on each infant, the generalized estimating equations of Zeger and Liang (20) were used to calculate odds ratios for all RI, all LRI, and LRI with wheeze. Comparing the usual logistic and generalized estimating equation approaches showed that estimates of effect were comparable with the two methods, but that most standard errors were slightly larger with the generalized estimating equation approach. Fixed covariates included birth order, sex, ethnicity, breast feeding to 6 mo of age, parental asthma/atopic status, household income, and crowding index. Covariates that change with time included age of the infant, calendar month, day-care attendance, and contact with a sick family member. All analyses were performed using the SAS data management and statistical analyses software (SAS Institute, Cary, NC, 1990).

The relative risk for the development of asthma was calculated using Cox's proportional hazards model. For this model the covariates included sex, ethnicity, breast feeding in the first 6 mo, income, parental history of atopy/asthma, and the number of nonwheezing illnesses prior to the diagnosis of asthma. Because the actual number of respiratory illnesses was included in this model, the variables associated with transmission of infection were omitted.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 1,315 infants were enrolled. However, 110 completed 30 d or less of observation and were excluded from the analysis. Moving out of the study area and placement of the child in day care for more than 20 h per week were the most frequent explanations for losses to observation. In addition, 11 subjects did not have a recorded parental history and were removed from the analysis. Of the 1,193 remaining infants, 835 (70.3%) were followed for the entire 18 mo. Follow-up to 18 mo was achieved for 69.7% for those without a parental history of atopy/asthma, for 68.8% of those with a parental history of atopy, and for 75.9% of those with a parental history of asthma.

The characteristics of the study population are summarized in Table 1. Slightly more than half were male and 38% were Hispanic. There was a parental history of atopy in 34% of the infants and a parental history of asthma in at least one parent in 16.4%. The majority had siblings, and 22.8% were fully breast fed during the first 6 mo of life. Fewer than 10% of the families earned less than $10,000 per year. More than two thirds of the infants attended day care for more than 5 h a day at some time during the follow-up. Division of the children into groups defined by parental history showed that those with a positive parental history included more infants who were fully breast-fed and more who attended day care for > 5 h/wk. However, the differences between the groups in the amount of time spent in day care did not quite reach statistical significance (p = 0.055). Lastly, the families with a positive parental history reported more 2-wk periods during which members of the family other than the proband were ill. The figures in the table refer to the percentage of surveillance calls in which the subject responded affirmatively to the question about contact with a sick family member.

The rates of all illness categories varied significantly by the age of the infant (all RI, p < 0.001; URI, p = 0.011; LRI, p < 0.001; LRI with wheeze, p = 0.019). The rates of all RI, URI, and LRI increased from birth, reaching a maximum at 7 to 12 mo of age and then decreased from 12 to 18 months of age (Table 2). The rate of LRI with wheeze was constant from 0 to 9 mo of age and decreased consistently thereafter. As an indication of the severity of these illnesses, 19.8% of the infants with a URI were taken to the doctor, whereas 43.3% of the infants with LRI and wet cough and 62.2% of the infants with wheezing were brought to medical attention.

The rates of all RI, URI, and LRI by parental history of asthma or atopy are shown in Figure 1. Rates for all three illness categories varied significantly by parental history. Rates were higher in infants with a parental history of atopy or asthma compared with infants without a parental history (all RI, p = 0.001; URI, p = 0.003; LRI, p < 0.007), and the illness rates did not differ significantly for infants with a parental history of atopy alone compared with those with a parental history of asthma. LRI rates were significantly higher in those with a parental history of asthma compared with those with a parental history of atopy alone (p = 0.007). The test of interaction between age and parental history was significant for LRI (p = 0.046), but not for all RI (p = 0.169) or URI (p = 0.098). The pattern of the rates shows that infants with a parental history of asthma had higher rates of LRI after 6 mo of age but not at younger ages.

View larger version (15K): [in this window] [in a new window]   Figure 1.   Rate of all respiratory illnesses (All RI), upper respiratory illnesses (URI), and lower respiratory illnesses (LRI) related to age and parental history. Parental history of asthma (closed squares); parental history of atopy (open circles); no parental history of atopy or asthma (dashed lines).

Rates of LRI with wheeze by age and parental history are shown in Figure 2, top panel. Illness rates varied significantly by parental history. Infants with a parental history of atopy or asthma had higher rates than did infants with no parental history (p < 0.001), and infants with a parental history of asthma had higher rates than did infants with a parental history of atopy alone (p = 0.004). These differences were most apparent after 8 mo of age, but the test of interaction between age and family history was not significant (p = 0.063). If the rate is expressed as a proportion of all LRI, this index of wheezing illness is maximal in the youngest infants and decreases consistently thereafter (Figure 2, bottom panel ). The infants with a parental history of atopy or asthma follow a pattern that is similar to those with a negative history, but with slightly higher rates and more variability.

View larger version (17K): [in this window] [in a new window]   Figure 2.   (Top panel ) Rate of lower respiratory illnesses with wheeze related to age and parental history. (Bottom panel ) Rate of lower respiratory illnesses with wheeze as a proportion of all lower respiratory illnesses. Symbols as in Figure 1.

Multivariate models using generalize estimating equations and including only parental history and age indicated that a parental history of asthma was significantly associated with the incidence of All RI (OR = 1.18, CI = 1.05 to 1.33), LRI (OR = 1.26, CI = 1.08 to 1.48), and LRI with wheeze (OR = 1.71, CI = 1.20 to 2.44). A parental history of atopy alone was significantly associated with the incidence of All RI (OR = 1.16, CI = 1.06 to 1.28), but not significantly with LRI (OR = 1.10, CI = 0.97 to 1.24) and LRI with wheeze (OR = 1.23, CI = 0.93 to 1.63). Removal of all the infants who ever received a diagnosis of asthma during the follow-up period did not change the results (n = 36).

To determine if the elevated rates of respiratory illnesses that were associated with a parental history of atopy/asthma could be explained by confounding variables and by other factors in the causal pathway, we modeled rates of respiratory illnesses based on variables known to be associated with increased transmission of infectious diseases. Models were built by first including age, sex, ethnicity, breast feeding, and income. Then exposure variables, which included month of illness, day care, siblings, crowding, and other ill family members, were added. Finally the parental history was added. For each outcome, addition of the exposure variables had the effect of slightly lowering the odds ratios for age. The odds ratios for sex, ethnicity, breast feeding, and income were essentially unchanged by addition of the exposure variables or parental history. Attendance at day care for > 5 h/wk, having more than one older sibling, and winter month were each associated with increased risk of all of the outcomes. Odds ratios ranged from 1.32 (CI = 1.14 to 1.53) for the association between number of older siblings and LRI to 3.03 (CI = 2.76 to 3.31) for the association between occurrence in a winter month and LRI.

After adjusting for age, sex, ethnicity, breast feeding, family income, month of illness, attendance at day care, number of siblings, and crowding, a history of contact with a sick family member was associated with increased rates of all the illnesses (Table 3). However, in the infants younger than 6 mo of age, the association did not reach statistical significance for the outcome of All RI. A parental history of asthma was associated with increased illness rates in the older, but not the younger, infants. A parental history of atopy was associated only with All RI. However, it should be noted that wheezing events were substantially less frequent than the other outcomes, which resulted in wide confidence intervals. Using LRI with wheeze as the outcome, a parental history of atopy had an odds ratio of 1.41 (CI = 0.98 to 2.03), which was higher that the odds ratio for the association between a parental history of atopy and LRI (OR = 1.06, CI = 0.92 to 1.22) or for the association with all RI (odds ratio = 1.14, CI = 1.03 to 1.26).

Asthma was diagnosed in 36 of the infants by the age of 18 mo. We modeled the risk of asthma using Cox's proportional hazards model with covariates for family history, age, sex, ethnicity, income, breast feeding in the first 6 mo of life, number of prior nonwheezing respiratory illnesses, and number of prior wheezing LRI. Only a family history of asthma and prior wheezing LRI were significantly associated with an increased risk of asthma. The relative risk for diagnosis of asthma in infants with a family history of asthma was 3.2 (CI = 1.5 to 6.9) compared with those with no family history of atopy or asthma. Infants with one prior wheezing LRI had an increased risk for diagnosis of asthma (RR = 7.2, CI = 3.2 to 16.1) compared with infants with no prior wheezing LRI. Infants with two or more prior wheezing LRI had a relative risk of 17.1 (CI = 7.0 to 41.9) compared with infants with no prior wheezing LRI. Nonwheezing LRI was also associated with an increased risk of asthma, although the increase did not approach statistical significance. Two to seven prior episodes of nonwheezing LRI resulted in a RR of 1.1 (CI = 0.4 to 2.9), whereas eight prior episodes resulted in a RR of 1.4 (CI = 0.3 to 5.7).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This analysis is based on a comprehensive respiratory illness surveillance system which documented events that required medical attention as well as those that did not. Consequently, we were able to enumerate mild upper respiratory illnesses and to be reasonably sure that all lower respiratory involvement was identified. Surveillance methodologies that quantify doctor-diagnosed disease are likely to select a sicker group of infants, even if subjects are encouraged to seek frequent care through a prepaid health plan (21). Our high illness rate compared with previous population studies (7, 21) is probably related to both the comprehensive nature of the follow-up and to our necessarily broad definition of lower RI.

We interpreted the report of two consecutive days of any combination of runny or stuffy nose, dry cough and trouble breathing as indicating respiratory infection. This inference is based on the documented concurrence of epidemics of respiratory illness characterized by these symptoms and peaks of isolation of viral isolates (22). We did not require positive viral cultures or antibody rises because of the known lack of sensitivity of these techniques (23), and the difficulty of applying them in a large population. The association of increased risk for each of the respiratory syndromes with variables known to increase transmission of respiratory infectionswinter season (22), attendance at day care (24), and older siblings (25)also supports this inference.

The key finding of this analysis is that a parental history of asthma increased the risk of respiratory illnesses, including mild URI in infants between 7 and 18 mo of age. These infants also had an increased incidence of all LRI and LRI with wheeze, whereas infants with a parental history of atopy had only an increased risk in the overall rate of RI (Table 3). Adjusting for factors associated with increased transmission of infection and removal of infants with diagnosed asthma did not change the findings. The latter analysis was performed to remove infants with established lung disease because this group has previously been shown to be at increased risk of viral respiratory infection (5, 12).

Previous studies of infants with atopy not associated with asthma have suggested that this group is not at increased risk of infection. For instance, Stempel and colleagues (13) followed 24 children who were attending day care together. Routine viral cultures were obtained and the results were compared with the incidence of wheezing illness and with IgE levels. Although elevated levels of IgE were associated with an increased incidence of wheezing illnesses, they were not associated with an increase in viral isolation. Cogswell and colleagues (14) selected infants at birth on the basis of a positive parental history of atopy or asthma. They found similar rates of viral isolation in infants who developed eczema or a positive skin prick test compared with infants who did not manifest atopy. Issacs and colleagues (15) studied children presenting with recurrent respiratory infections and their unaffected siblings. They found that clinically diagnosed respiratory infection and virus isolation was independent of skin test positivity and serum IgE.

The difference between these earlier studies and the current investigation can be explained, in part, by differences in study design. We compared infants with a parental history of atopy with those with no parental history. The studies of Cogswell and Issacs and their colleagues used siblings or unrelated infants with a similar parental history as control subjects, so they could not have detected differences between children with a positive parental history and those with a negative parental history. The results of Stempel and colleagues (13) are based on observations of only 24 children. Although they were followed for 5 yr, only 576 URI and 59 LRI were observed. This is probably too few events to detect differences of the magnitude that we observed. More important, the children were classified as allergic on the basis of IgE alone. The genetic predisposition to allergy that accompanies a positive parental history is probably more complex than abnormal regulation of a single immunoglobulin, so the results of the two studies are not necessarily incompatible with one another.

We predicted that there would be a difference in the illness rate between infants younger and older than 6 mo of age because the first 6 mo of life are critical in the development of the immune system. Lymphocyte, macrophages, and pulmonary dendritic cells are all hyporesponsive in the neonatal period (26). Cord blood contains less than 10% memory T-cells compared with 50% in adults, and cytotoxic T-cells are less active in neonates (26). In addition, infants younger than 8 mo of age have less of an IgA response to infection with RSV than infants 9 to 21 mo of age (27), and the tendency for recurrent infection has been correlated with low convalescent levels of IgA (28). Thus, the determinants of infection and recurrence are markedly different in very young infants compared with older ones.

The first 6 mo of life may also be important in determining the outcome of exposure to allergens. During normal development, newborns respond to allergen by producing both IgG and IgE antibodies (29). In most children the IgE component is shut down in later infancy, but in those with an atopic predisposition, allergen-specific Th₂ responses (production of IL-4 and IgE instead of IL-2 and INF, which is produced with a Th₁ response) may persist (30). It has been suggested that viral infection during a critical period of development could determine if subsequent immune responses are characterized as Th₁ or Th₂ (11). Although most viruses stimulate the production of interferon- and suppressor CD8 cells, both of which would tend to suppress the Th₂ response (31), RSV infection produces RSV specific IgE (32), which could be an indication of activation of the Th₂ response. Thus, mechanistically viral infection could facilitate or suppress the development of allergic responses.

Our finding that the respiratory illnesses that occur in the first 6 mo of life are not associated with either a parental history of allergy or asthma is in agreement with those who have argued that wheezing in very young infants is not related to the development of asthma (6). However, these investigators include infants as much as several years of age in their categorization of "young," whereas we found evidence of an effect of parental history in infants as young as 7 mo. We, therefore, suggest that the reaginic immune system is active enough by this age to be responsible for some of the manifestations of respiratory illnesses in this age group. This is biologically plausible since IgE specific for inhalant antigens has been demonstrated in the blood of infants as young as 5 mo of age (29). On the other hand, our finding that the rate of wheezing and the proportion of LRI with wheeze decreases consistently throughout infancy in normal infants is compatible with the notion that wheezing in this group is related to small airways in the very young infants (33).

While a major strength of this study was its detailed prospective follow-up of the infants, one probably unavoidable weakness in a study of this size is the inability to intensively evaluate each subject for each illness. Diagnostic misclassification is inevitable, and given the current state of technology it would be impossible to precisely localize the sites of involvement within the respiratory tract without invasive procedures. By reporting the total event rate and the subcategories of LRI and LRI with wheeze we have attempted to provide a comprehensive description of clinically recognizable respiratory syndromes. A shift in the distribution between LRI with and without wheeze would not have changed the major conclusions of the analysis since a parental history of asthma was important for both outcomes.

Another potential source of misclassification is in the interpretation of parental history for atopy/asthma as an indication of a genetic predisposition to allergy/asthma (34). Although we and others have shown the importance of a parental history as a predictor of disease (3, 25), this is an imprecise way to measure genetic influences. Not every child in the family inherits the allergic tendency, and considering them all together is bound to bias the results towards the null. We assume that the genetic effects are actually stronger than demonstrated by this type of classification.

In conclusion, we have demonstrated that a parental history of asthma is a risk factor for respiratory illnesses in both the upper and lower respiratory tract. Although respiratory infection is an important cause of morbidity in its own right, the significance of our findings depends, in part, on the relationship between these episodes and the development and expression of asthma. Because asthma was diagnosed infrequently we did not have sufficient power to rule out an effect of nonwheezing LRI on the development of asthma. However, the risk of developing asthma was not lower in those with frequent nonwheezing LRIs than in those without, as would be expected if viral infections protect against the development of asthma. Thus, elucidating the relationship between viral infection and the development of asthma remains an important area for research.