INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Wheezing in infancy is very common, 34% of infants experiencing at least one wheezing episode by their third birthday, and 49% by their sixth birthday (1). However, not all children who experience early life wheezing continue to wheeze. Approximately one half of those wheezing before 3 yr of age are asymptomatic by 6 yr of age, at which time the prevalence of current wheeze is 29%. The prevalence of doctor-diagnosed asthma is even lower: at 7 yr of age, asthma has been diagnosed in 12% of children, and a further 5% have recurrent wheezing but have not received a diagnosis of asthma (2). The proportion of wheezing infants who do not develop ongoing asthma varies between studies, depending on the size and type of cohort and the length of follow-up, but it is at least 50% (3).

Therapeutic decisions should be made in the light of a firm understanding of the pathophysiology and the natural history of the disease in question. If all children presenting with wheeze were to be treated with antiasthma medications, we would need to treat half of all children. This would be neither feasible, being logistically difficult and very costly, nor desirable, as clearly the majority will not develop asthma (1, 4, 5). At present there is no way of identifying individual infants who are at high-risk of continuing respiratory symptoms as they grow older so that they could be targeted for appropriate preventative treatment. It has been suggested that both inhaled corticosteroids and sodium cromoglycate not only reduce morbidity and improve quality of life in the short term, but lead to an amelioration of future disease (6). However, inhaled corticosteroid preparations have the potential for deleterious long-term side effects, especially when given at a time of rapid lung growth, and cromoglycate must be given three to four times per day to be effective. Thus, to justify a treatment regimen that includes these agents, indices that allow a more accurate prognosis of wheezing in infancy are needed (8).

Data from a number of longitudinal studies of wheezing in childhood suggest that atopy, and particularly eczema, is the single most important risk factor in the development of wheezing illness and its persistence (2, 9). In addition, susceptibility to wheezing and to asthma seem to be determined by certain factors, exposures, or events that exert their influence in very early life, perhaps even before birth, and these include fetal nutrition (13, 14), duration of pregnancy (15), tobacco smoke exposure (16, 17), environmental air pollution (18), postnatal nutrition, breast feeding, family size (19, 20), maternal age (5, 21) socioeconomic status (5), and allergen exposure (22). By paying attention to such factors, it becomes possible to identify children who have at least a 50% chance of developing asthma in the future.

The role of T-cells and their cytokines in the regulation of IgE-mediated hypersensitivity to allergen is a current focus of major interest. Two distinct populations of T-helper CD4+ lymphocytes exist in the mouse, Th-1 cells producing interferon- (IFN-) and Th-2 cells producing interleukins 4, 5, 10, and 13. Th-1-like and Th-2-like T-cells also appear to exist in the human (23). Th-2 cytokines promote the mast cell and eosinophil-mediated inflammatory response, and the isotype switching of B-cells to IgE (24), whereas the Th-1 cytokines are linked with the classic delayed hypersensitivity response. It appears that repeated episodes of environmental stimulation during early life cause one of these two T-cell phenotypes to become dominant, directing the immune response in later life (25). Variations in the rate of T-cell maturation may contribute to the development of different patterns of immunologic responsiveness to environmental allergens (26). A number of recent studies have demonstrated quantifiable differences in markers of T-cell activation and antigenic priming (27), cytokine production (28), and proliferative responses to allergen (29) in cord blood between infants who go on to develop atopic eczema compared with those who do not. Soluble IL-2 receptor (sIL-2R) is raised in children with asthma and appears not to be affected by corticosteroid treatment despite clinical improvement (30). In addition, Leonard and colleagues (31) have demonstrated reduced production of IFN- by atopic asthmatic subjects in response to Dermatophagoides pteronyssinus when compared with nonatopic control subjects. Thus, it is possible that these indices may have a role in predicting the persistence of wheezing illness in infants who have already begun to wheeze, and could be used to improve our ability to formulate a prognosis for individual wheezing infants.

In order to determine which factors or combination of factors might be predictive for the persistence of asthma symptoms, we undertook a prospective longitudinal study of a group of infants with recent onset wheezing. This report presents our findings relating to demographic and immunologic markers and their usefulness as predictors of ongoing wheeze requiring treatment 1 yr after presentation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

One hundred nine children were entered into the study. Inclusion criteria were: age, 3 mo to 3 yr; recent onset (in the previous 12 wk) of wheeze (either one major episode lasting at least 24 h or three minor episodes lasting at least 6 h); and at least one parent having a past or present history of asthma or eczema and current positive skin prick test responses to a standard panel of allergens. Children were excluded from the study if their first episode of wheeze was secondary to documented or suspected respiratory syncytial virus (RSV) bronchiolitis. Children suitable for recruitment were identified by their Primary Care Physician or after attendance at the in-patient and out- patient departments of Southampton General Hospital. The parents were given a written information sheet and invited to join the study, families being given as much time as required to make their decision. Written parental consent was obtained, and the study was performed with the approval of the Southampton University and Hospitals Ethical Subcommittee. Recruitment to the study took place between October 1993 and March 1995. Each child was followed initially for 12 mo, although the study of this cohort is ongoing.

Clinical Assessment

At the first study visit, a full history was taken from the parents, including demographic details, family history of atopic disease, details of the home, including types of heating, presence of pets, number, ages, and smoking history of all occupants, and birth weight, feeding history, and full medical history of the index child, including details of all respiratory symptoms to date. The child was weighed (Seca 835 electronic scales; Seca Ltd., Birmingham, UK) and heighted using supine length until the age of 24 mo (Kiddimetre; Raven Equipment Ltd., Dunmow, Essex, UK) and standing thereafter (Invicta Plastics Ltd., Leicester, UK), and values were plotted on a centile chart (Tanner-Whitehouse 1970 Integrated Series). A full clinical examination was performed by the same clinician for all children (J.B.C.). Eczema was diagnosed according to the criteria of Hanifin and Rajka (32), namely, erythematous papular, vesicular chronic skin eruptions with evidence of itchiness lasting for longer than 4 wk, and affecting either face (excluding nose and lips), torso, and extensor surfaces of limbs or skin flexures.

A daily record of each subject's respiratory symptoms (cough and wheeze), antiasthma treatment, and life events was kept by the parent for 12 mo using patient diary cards. These were personalized for each child and were designed to be quick and easy to complete.

Allergy Skin-prick Testing

Each study child and both parents were skin-prick tested at entry to the study using a positive histamine control (1%), negative control (phenol/saline, 0.9%), grass mix (2.5%), cat dander (150%), dog dander (150%), house dust mite (Dermatophagoides pteronyssinus) (1.2%), cow's milk (50%), egg white (6%), and Aspergillus fumigatus (5%) (SmithKline Beecham, Welwyn Garden City, UK). A weal size of  2 mm in the absence of a reaction to the saline control was regarded as a positive response. Children receiving antihistamine preparations at entry to the study were not skin-tested.

Blood Samples

Local anaesthetic cream (EMLA; Astra Pharmaceuticals Ltd., Kings Langley, UK) was applied to both antecubital fossae of each subject on arrival for Visit 1. A blood sample of 8 ml was obtained, 3 ml being placed into heparin and 5 ml into clot-activating tubes. Serum was separated between 1 and 2 h after sampling by centrifugation at 2,500 rpm for 10 min and was stored at 70° C until analysis.

Soluble IL-2 Receptor

Soluble IL-2R measurement was performed using the Cellfree ELISA Test Kit (T Cell Sciences, Inc., Cambridge, MA) as described previously (33).

Proliferation Assays and Assay for Interferon- Production

Isolation of peripheral blood mononuclear cells. Peripheral blood mononuclear cells (PBMC) were isolated from heparinized venous blood by density gradient centrifugation on lymphocyte separation medium (ICN Biomedicals, Irvine, Scotland) and washed in culture medium (RPMI 1640 with 2 g/L sodium bicarbonate supplemented with 2 nmol/ L L-glutamine, 100 µg/ml streptomycin, and 100 U/ml penicillin).

Proliferation assays. Peripheral blood mononuclear cells were cultured with or without stimulants as described previously (28). Final concentrations of stimulants were as follows: PHA, 1 µg/ml (Sigma, Poole, UK); anti-CD3 antibody, 50 ng/ml (Ortho Diagnostics, Raritan, NJ); house dust mite, 2,000 U/ml (National Institute for Biological Standards and Control - First International Standard, Potters Bar, UK); cat fur extract, 2,000 U/ml (NIBSC); and betalactoglobulin, 10 µg/ ml (Sigma). These concentrations were optimized as described previously (28). Mitogenic responses to antigens were investigated by stimulating the PBMCs from five blood samples with the optimal doses of allergen for both 3 and 6 d. No positive proliferative responses ( 2 × background) were demonstrated in the 3-d cultures, indicating that the allergens did not have a mitogenic effect. Results were expressed in counts per minute as a ratio of incorporated [³H]thymidine of stimulated cells to unstimulated cells.

Preparation of PBMC supernatants for IFN- assay. PBMCs were suspended at 1 × 10⁶ cells/ml in culture medium supplemented with 10% filtered autologous plasma in 24-well microtitre plates (Costar, Cambridge, MA) with a total volume of 2 ml/well. Cells were stimulated with anti-CD3 (50 ng/ml), cat fur extract (2,000 U/ml), BLG (10 µg/ml), or house dust mite (2,000 U/ml) (National Institute for Biological Standards and Control) or incubated in medium alone for 24 h. The supernatants were then harvested and stored at 80° C. This time period was utilized rather than 48 h as described in (34) as these supernatants were also analyzed for IL-2, IL-4, and IL-5 for which the optimal harvesting time is 24 h (35) and after which time point there is a dramatic reduction in supernatant concentrations of these cytokines (D. J. Quint, Ph.D. thesis). IFN-, however, showed no significant increase or decrease between 24 and 48 h.

Assay for IFN- production. IFN- concentrations were determined by use of an ELISA for human IFN- in medium (Holland Biotechnology, Leiden, The Netherlands) with a minimum detection level of 10 pg/ml. The ELISA was carried out according to manufacturer's instructions.

A number of further laboratory investigations were performed, but as the data were not used in the statistical analysis these will not be discussed here.

Clinical Follow-up

Arrangements were made for each family to attend at a time of their choosing every 3 mo (a total of five visits) for clinical assessment, and a phone call was made to each family midway between visits. Additionally, parents were encouraged to bring the infants for extra assessments if there was any persistent skin, bowel, or chest problem. At each visit, information relating to changes in the household situation (including new pets), current and new symptoms, and current medication was recorded. Completed diary cards were collected and checked, and new cards given. During the follow-up period, clinical management was the responsibility of the child's Primary Care Physician and the investigators ensured that this was according to the International Paediatric Asthma Consensus Group Guidelines on asthma management (36), which were circulated to all Primary Care Physicians at the commencement of the study. These guidelines recommend dividing patients by severity into those with mild infrequent episodic disease (treatment: bronchodilators alone), frequent episodic disease (treatment: bronchodilators, plus prophylaxis with ketotifen, sodium cromoglycate, or low dose inhaled corticosteroids), and chronic persistent disease (treatment: bronchodilators, plus inhaled corticosteroids). At each scheduled visit the principal investigator recorded the treatment given.

Outcome Measures

On completion of the 12-mo study period, it was noted whether each child was currently receiving prophylactic antiasthma treatment (oral ketotifen, inhaled sodium cromoglycate, or inhaled steroids). If so, they were regarded as having significant persisting disease. In addition, the number of days during the final 90 d of the study that each child experienced symptoms was calculated. One subject was excluded from analysis because of the diagnosis of hyperimmunoglobulin E syndrome.

Statistical Methods

All data were entered onto a dedicated database and checked for errors. Descriptive statistics were determined. Ten variables were chosen as being of special interest: (1) atopy in the infant: i.e., serum IgE > 1 SD above age-related normal (Dr. F. Stevenson, personal communication) and/or eczema at recruitment and/or positive allergen skin test; (2) degree of parental atopy: paternal or maternal or biparental atopy; (3) number of older siblings; (4) age at presentation with wheeze; (5) sex; (6) serum soluble IL-2R; (7) proliferation of peripheral blood mononuclear cells (PBMC) to -lactoglobulin; (8) proliferation of PBMC to cat; (9) production of IFN- on stimulation of PBMC with -lactoglobulin; (10) production of IFN- on stimulation of PBMC with D. pteronyssinus.

Each variable was entered separately into univariate models. The goodness of fit of these models was assessed using log likelihood, which followed a chi-square distribution if the null hypothesis was true and the sample size large. Estimated logistic regression coefficients, standard errors, odds ratios, and p values from the score test statistic on the significance for each variable's inclusion into each univariate model were calculated.

The predictive accuracy of the univariate logistic regression models was calculated using the following parameters.

1. 1. Percentage of subjects correctly predicted symptomatic: the number of infants predicted as having a positive outcome divided by the number actually observed as having a positive outcome.
2. 2. Percentage of subjects correctly predicted asymptomatic: the number of infants predicted as having a negative outcome divided by the number actually observed as having a negative outcome.
3. 3. The overall number incorrectly predicted by the model.
4. 4. The overall number correctly predicted by the model.

Using multiple regression analysis, a step-down model (variables tested and dropped from the full model) and a step-up model (variables entered one by one and tested for inclusion) were used. A variable was entered if the score test had a significance level of p < 0.05 and was removed if the likelihood ratio significance was p > 0.1. As only four infants had data on all 10 variables, the full model was not appropriate, and additional models were fitted using the same methods.

Six multivariate models were selected for further investigation. The number of infants in each model and the predictive accuracy of each model were assessed, and model diagnostics were performed. The six multivariate models that were selected for further investigation were: (1) infant atopy + constant; (2) infant atopy + age + soluble IL-2R + interferon gamma production to house dust mite + constant; (3) infant atopy + age + soluble IL-2R + sex + constant; (4) infant atopy + age + soluble IL-2R + both parents atopic + constant; (5) age + soluble IL-2R + constant; (6) age + soluble IL-2R + age/ soluble IL-2R interaction + constant.

The numbers of infants included in each model, the overall percentage of the model predicting correctly, the sensitivity, the specificity, the positive predictive value, the negative predictive value, and the number of infants who did not receive treatment but who were predicted to receive treatment by the model with a greater than 0.5 probability were calculated.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

One hundred seven subjects completed the study, of whom 78 (73%) were boys and 29 (27%) were girls. Mean age at recruitment was 10.9 mo (SD, 7.36; median, 9.0 mo; mode, 5.0 mo; range, 3.0 to 35.0 mo).

A positive clinical outcome, regarded as a child requiring prophylactic antiasthma treatment 1 year after recruitment, was observed in 53 (49.5%) children. Of these, 12 received ketotifen alone, two received ketotifen in conjunction with sodium cromoglycate, and one received ketotifen in conjunction with inhaled corticosteroids. A further three received sodium cromoglycate and 35 inhaled corticosteroids. The majority (35; 66.0%) of the children receiving treatment were experiencing frequent symptoms, having been symptomatic on at least 30 of the last 90 d of the study. Of the 54 not receiving treatment, 15 (27.7%) had symptoms on  30 of the last 90 d of the study.

Personal History

Fifty (46.7%) subjects had no older siblings, 36 (33.6%) had one, and 21 (19.6%) had two or more; 39 (36.4%) infants were never breast fed, 12 (11.2%) were breast fed for less than 1 mo, and 54 (50.5%) were breast fed for 1 mo or more (two subjects unknown). Fifty-seven (53.3%) subjects had suffered eczema before enrollment.

Skin-prick Testing

At recruitment 11 subjects were receiving treatment with ketotifen and were not therefore skin tested. Of those who were, 27 (28.1%) infants were skin-prick test positive and 69 (71.9%) were negative. The allergen giving rise to the greatest number of positive responses was egg white (20 children), followed by cat (15), house dust mite (14), and grass (seven). Results to individual allergens are shown in Table 1.

Laboratory Investigations

Serum assays. Mean total serum IgE for the group as a whole was 46.8 IU/ml (SD, 196.9; median, 0; range, 0 to 1,900.0), reference ranges being: 0 to 12 mo < 11 IU/ml; 13 to 24 mo < 29.2 IU/ml; 25 to 36 mo < 42.5 IU/ml. The mean was elevated by a small number of individuals having very high values. Mean sIL-2R was 1,504.3 ng/ml (SD, 933.2; median, 1,379.6; range, 281.1 to 7,924.8). Median (range, SD) published values in healthy nonasthmatic, nonallergic children of this age range are 1,355 (965 to 2,570, 466) ng/ml (33).

Proliferation assays. The results of stimulation of peripheral blood mononuclear cells are shown in Table 2.

IFN- Production

The concentration of IFN- produced by PBMC on stimulation is shown in Table 3. The maximum response of 44,439.30 pg/ml to stimulation by PHA was large compared with the other responses. The next largest value was 23,821.70 pg/ml. The two largest responses of 23,288.5 and 20,612.4 pg/ml to stimulation by D. pteronyssinus were very large compared with the other responses. The next largest response was 3,639.6 pg/ml.

Univariate Logistic Regression Analysis

Using univariate logistic regression, we found that children were more likely to experience persistent symptoms and to require regular treatment if, at presentation, they were atopic with eczema and/or positive skin-prick tests and/or raised serum IgE (p = 0.0025, OR = 3.58), were older (p = 0.018, OR = 1.07), and if both parents were atopic (p = 0.33, OR = 1.58). Results of univariate analysis are given in Tables 4 and 5.

Multivariate Analysis

In Table 6, the following are shown for each of the six multiple logistic regression models investigated: the numbers of infants in whom the relevant data were available to enable their inclusion in the model; the ability of the model to predict clinical outcome correctly overall (%); the sensitivity; the specificity; the positive predictive value; the negative predictive value; and the number of infants who were asymptomatic but were wrongly predicted by the model as being symptomatic with a greater than 0.5 probability. This last is of particular interest when planning intervention studies as it reflects the number of children who would have received treatment inappropriately had the model been used to direct management. All six models fitted the data reasonably well. Using information on the predictive accuracy of each model, the statistically superior model was found to be Model 6, which combined information on age at presentation, the concentration in the serum of soluble interleukin-2 receptor, and the interaction between the two. This model was able to include 97 of the 107 children completing the study, and it was 67% accurate in predicting the clinical outcome overall, correctly identifying 61% of symptomatic and 73% of the asymptomatic children with a true positive rate of 67% and a true negative value of 67%. Using this model, the risk of a child continuing to wheeze can be assessed with only a 14.4% chance of wrongly predicting as being symptomatic a child destined to become asymptomatic.

However, for the purposes of appropriate recruitment to an intervention study, Model 5, which combined information on age at presentation and the concentration in the serum of soluble interleukin-2 receptor but omitted the interaction between the two, would seem to be preferable. For the same proportion of children, it allowed 71% accuracy in predicting the clinical outcome overall, correctly identified 57% of symptomatic and 84% of the asymptomatic children with a true positive rate of 76% and a true negative value of 68%. Using this model, the risk of a child continuing to wheeze can be assessed with only an 8.2% chance of wrongly predicting as being symptomatic a child destined to become asymptomatic. This model has the added advantage of simplicity.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Evidence that early intervention with anti-inflammatory agents leads to an amelioration of future disease is increasing, and the demand for rational treatment strategies for wheezing in infancy is growing. Thus, the accurate selection of subjects in whom the benefits of early intervention outweigh the disadvantages has become a most important issue.

Using univariate logistic regression, we found that children were more likely to experience persistent symptoms and to require regular treatment if, at presentation, they were atopic with eczema and/or positive skin-prick tests and/or raised total serum IgE, were older at presentation, and if both parents were atopic. Using multivariate logistic regression analysis, we have been able to describe a simple model for the prediction of disease requiring treatment that persists for at least 12 mo after presentation in infants and young children already at high risk of developing atopic disease by nature of their family history. By recording the age of first wheezing and by obtaining a 2-ml sample of venous blood for the measurement of sIL-2R, the risk of a child continuing to wheeze can be assessed with only an 8.2% chance of wrongly predicting as being symptomatic a child destined to become asymptomatic. In contrast, by using a first-degree family history of atopic disease alone, as for recruitment in this study, it was only possible to predict persistent symptoms with at best a 50% accuracy, as demonstrated by the clinical outcome in our group of patients. The aim of this study has therefore been met: to provide a tool for the rational recruitment of young children with wheeze to future intervention studies. Although at the present time the most commonly used antiasthma prophylactic agents are inhaled corticosteroids, the discipline of pharmacotherapy is far from static and it is likely that new agents and new indications for existing classes of drugs will emerge. In this study, the usage of prophylactic agents was pragmatic and based on guidelines of current clinical practice. Future randomized controlled trials need not be limited to the drugs used for treatment in our study group, but rather should include the most effective agent available at the time.

All children included in this study had at least one atopic parent, which does, to a certain extent, limit the generalizability of these data. However, the German Multi-centre Allergy Study (MAS) indicates that 50% of all children who ultimately become asthmatic derive from such families, and therefore our observations are applicable to half of all asthmatic children. The size, and also the cost, of this study would have been increased by a factor of 4 to 5 if based on an unselected population. In addition, venepuncture might not have been considered ethical in a group of children without a strong atopic family history. We therefore thought it appropriate in the first instance to conduct a study on a high yield population. If positive associations were revealed, an extension of the study to a general population sample would be justified. If the results were negative, the likelihood of a whole population sample yielding different results would be small. We would now strongly support this study being repeated using a study group drawn from a general population sample.

In this long and relatively onerous study only two out of 109 subjects were lost to follow-up. This high compliance was achieved by a number of strategies. We were at all times realistic about the amount of work this study involved for the families and made this clear at recruitment. However, the participating families showed great dedication to this study, which we believe to be a reflection of the degree of concern felt by these parents and the relative paucity of help and information available for families with a wheezing infant. The very low dropout rate renders these data more robust and increases the value of the study.

Subjects were recruited to this study within 12 wk of first ever wheezing because in future intervention studies employing treatment with anti-inflammatory agents, the aim will be to commence treatment as early as possible so as to treat before airway inflammation becomes fully established. This led to stringent recruitment criteria that might have proved difficult to meet. However, we received considerable support from the local Primary Care Physicians, who find the management of these wheezing infants increasingly problematic, and from the parents, many of whom already had an older child with wheezing illness and who were therefore eager to contribute to a study relating to this subject.

Although follow-up of this group continues and the whole group will be reevaluated clinically at 6 yr of age, the length of follow-up reported here is 12 mo. This period was chosen as representing a degree of disease for which treatment was considered to be indicated in children with a strong family history of atopy and allergic disease. If, at the commencement of wheezing in such children, it could be predicted that disease would continue for 12 mo, the majority of physicians would regard prophylactic antiasthma treatment as being appropriate. Although a robust diagnosis of asthma will not be possible until 6 yr of age or older, there is a need to identify which children have ongoing disease necessitating therapy so that early intervention studies can be planned.

The markers, both clinical and laboratory, assessed in this investigation were chosen on the basis of published evidence supporting the role of each as a risk factor for the development and persistence of wheezing. Skin-prick testing to allergens was performed despite the low predicted prevalence of positive results in this age group as the presence of a positive response would imply significant atopy and associated persistence of symptoms (37). Unfortunately, this study had begun before the recent evidence for the importance of egg allergy in the development of asthma (38) was published, and egg was not included in the PBMC stimulations, although it was the commonest allergen to give a positive skin-prick test. Markers of eosinophil activation have been demonstrated to correlate with airway function in childhood and have been used as a measure of current eosinophil-activated inflammation in the airways (39), and were therefore measured in both serum and urine, although in the current analysis these did not prove to be prognostic. Recent studies show that IFN- and IL-4 reciprocally regulate IgE synthesis and the dominance of the Th-1 or Th-2 phenotype. As quantifiable differences between infants who develop atopic disease and those who do not have been demonstrated (28, 30), we therefore measured T-cell proliferative responses and IFN- production to allergen. Similar differences have been shown to extend to markers of T-cell activation and antigenic priming (27) and the soluble IL-2 receptor (30), and these indices were therefore also measured. Longer-term follow-up will allow the evaluation of the relevance of symptom phenotype to outcome. Clinically discrete phenotypes such as viral-induced wheezing may prove to demonstrate different natural histories when compared with ongoing atopic asthma.

Atopy in the infants themselves appears not to have had as great an effect as might have been expected on the clinical outcome, and this may be because all infants were selected on the basis of parental atopy. That birth order did not prove to be a significant risk factor may be explained by the small sample size and the relatively short period of follow-up. These children will be followed up regularly and reassessed at 6 yr of age.

The risk of demonstrating persistent wheeze in this group of children was increased with increasing age at presentation. Wilson and colleagues (40) have described a similar finding: wheezing in the first year of life was associated with increased bronchial responsiveness but a good prognosis, whereas later onset wheezing was associated with atopy and was more likely to be persistent. In the longer-term follow-up of the same group of children (41), parental atopy, and current personal atopy and bronchial hyperresponsiveness (BHR) were associated with both attack severity and interval symptoms. Although these children had been followed since 2 yr of age, no attempt has been made to examine the predictive value of atopy and BHR, or indeed any other indices, at 2 yr of age on symptoms at 6 yr of age.

The day-to-day clinical management of these children remained under the jurisdiction of their Primary Care Physicians as this was an observational study, and therefore the therapeutic decisions made were likely to reflect management in Primary Care. That the Primary Care Physician was aware that the child was under follow-up by a hospital specialist might have affected the threshold for the instigation of prophylactic treatment, but this effect could have acted in either direction. In all cases, management was according to the International Paediatric Asthma Consensus Group Guidelines on asthma management, and this might explain the relatively high incidence of treatment with antiasthma prophylaxis. In this age group, ketotifen is considered as a first-line option. Fifteen children were symptomatic during the last 90 d of the study but were not receiving prophylactic treatment. This may be a reflection of the time taken by Primary Care Physicians to make the decision to change management and to respond by commencing a new drug treatment once an increase in the frequency and severity of symptoms has been noted. A time lag between such symptoms and the issuing of a prescription is inevitable. However, these children may have experienced only temporary symptoms for which treatment was not indicated. As short-lived symptomatic episodes are common in early life, assessment at any time point will identify a small number of children who are symptomatic but not receiving treatment. Here we present our findings after a 1-yr follow-up period. Reassessment at 6 yr of age will allow us to reevaluate our model.

The major outcome measure employed was whether or not prophylactic treatment had been initiated within the 12-mo follow-up period. This was felt to be more stringent than the documentation of symptoms experienced by the subject for a number of reasons. Effective antiasthma therapy should diminish or eliminate symptoms, rendering symptom reporting unhelpful in the assessment of children receiving such treatment (50% of subjects), although the majority of these subjects continued to demonstrate wheeze, often because of the short duration of therapy prior to the end of the study. In this study, symptoms were recorded by a third party (the parent), which could have led to underreporting or, indeed, to overreporting. A considerable interparent difference in thresholds for symptom reporting exists. The month of recruitment and therefore the month in which follow-up ceased differed between individuals, and those children finishing in autumn and winter would be more likely to be symptomatic in the final months of follow-up because of the higher prevalence of viral upper respiratory tract infections. In order to minimize these factors, treatment history was used as the major outcome measure. The therapeutic agents used in these subjects appeared to represent the International Paediatric Asthma Consensus Group Guidelines on asthma management, the agents used being ketotifen, sodium cromoglycate, and inhaled corticosteroids. The relatively frequent usage of ketotifen reflects the young age and strong atopic family history of the study group.

In summary, we have examined the potential risk factors for wheeze that persists for at least 12 mo after presentation in a group of young children, each with at least one atopic parent, with early-life wheezing. We have shown that increased age at presentation, personal atopy, and raised soluble IL-2R are all associated with increased risk. In addition, using multivariate logistic regression analysis, we have described a simple model for the prediction of ongoing wheeze that can be applied within 12 wk of the commencement of symptoms. Whereas it is essential that these findings are replicated, we suggest that it may now be possible to predict which young children with wheeze will continue to wheeze for 1 yr after presentation.