Early Respiratory Infections and the Development of Asthma in the First 27 Years of Life

Aino K. Rantala; Maritta S. Jaakkola; Elina M. S. Mäkikyrö; Timo T. Hugg; Jouni J. K. Jaakkola


Am J Epidemiol. 2015;182(7):615-623. 

In This Article


Study Population

The source population included all children in the city of Espoo, Finland, who were born between January 1, 1984, and March 31, 1990. Espoo, located near the western border of Helsinki, is the second largest municipality in Finland, with a population of 265,646 in 2014.[20] A parent-administered baseline questionnaire was distributed in March 1991 to a random sample of children drawn from the roster of Finland's Statistical Center.[21] The baseline population included 2,568 children whose parents filled in the questionnaire. In March 1997, we conducted a 6-year follow-up survey of the cohort, with a follow-up rate of 77.3%. The characteristics of the baseline study population and the 6-year cohort have been described in detail elsewhere.[21–23] In 2010–2011, we conducted a 20-year follow-up and received 1,623 responses (follow-up rate = 63.2% of the baseline study population). Part 1 of the present analyses was conducted among a cohort of 2,228 subjects (86.8% of the baseline population) who had no asthma at baseline (81 subjects were excluded due to asthma) and who participated in at least 1 follow-up survey (259 were excluded due to nonparticipation). Part 2 was conducted among the whole baseline study population (n =2,568). Figure 1 presents data on participation in the different phases of the study. The study protocol was approved by the Ethics Committee of Oulu University Hospital (Oulu, Finland).

Figure 1.

Selection of subjects for the Espoo Cohort Study, 1991–2011. Part 1 of the current analyses included 2,228 subjects after exclusion of those with no follow-up data (n = 259; 10.1%) or with previous or current asthma reported at baseline (n = 81; 3.2%). Part 2 of our analyses was conducted in the whole study population (n = 2,568; 100%). A total of 1,984 (77.3%) subjects participated in the 6-year follow-up, and 1,623 (63.2%) participated in the 20-year follow-up. Altogether, 686 (26.7%) subjects participated only in the 6-year follow-up, 1,298 (50.5%) subjects participated in both follow-ups, and 325 (12.7%) participated only in the 20-year follow-up.

Health Outcome

The outcome of interest was the development of asthma during the follow-up period. In the baseline and 6-year follow-up surveys, this was determined with the question, "Has the child ever had physician-diagnosed asthma?" If the answer was yes, the age at which the asthma had started was requested. The following questions were asked in the 20-year follow-up: "Have you ever had physician-diagnosed asthma? If yes, at what age did it start?" We applied survival analysis in which we used the information on the age of asthma onset.

Respiratory Tract Infections as Determinants of Asthma

In part 1 of our analyses, the occurrence of respiratory infections during the 12 months preceding baseline data collection was assessed with the following question: "How often did the child experience the following infections during the past 12 months?" The list of infections included the common cold, tonsillitis, sinusitis, otitis media, acute bronchitis, and pneumonia. To judge the relative frequency of common cold and otitis media, we used the age-specific frequency distribution of each infection in the total study population in defining the exposure and reference categories (the cutoff point was at the 75th percentile). In children under 3 years of age, the exposure category included at least 4 infections and was coded 1, while fewer than 4 infections was used as the reference category (coded 0). For acute otitis media, the corresponding cutoff point was at least 2 infections (coded 1) versus fewer than 2 infections (coded 0). In the older children, the cutoff for common cold was at least 3 infections (coded 1) versus fewer than 3 infections (coded 0), and for acute otitis media it was at least 1 infection (coded 1) versus no infection (coded 0). For tonsillitis, sinusitis, acute bronchitis, and pneumonia, the cutoff point was at least 1 infection (coded 1) versus no infection (coded 0). Respiratory infections were classified as LRTIs (including acute bronchitis and pneumonia) or URTIs (including the common cold, tonsillitis, sinusitis, and otitis media).

Recurrent otitis media before the age of 3 years was defined as at least 3 episodes of acute otitis media during the past 12 months and/or a confirmatory answer to the question, "Has the child had recurrent otitis media?".

In part 2 of our analyses, information on respiratory infections leading to hospitalization from birth to age 6 years was retrieved from the National Hospital Discharge Register using computerized record linkage, with the Finnish personal identification number used as the key. The National Hospital Discharge Register includes the dates and causes of all hospitalizations (hospital admission requiring an overnight stay) of Finnish citizens that have occurred since January 1969. The diagnoses were coded according to the International Classification of Diseases, Eighth Revision, between 1969 and 1986 and the Finnish version of the International Classification of Diseases, Ninth Revision, between 1987 and 1995.

Respiratory infections requiring hospitalization included all of the main diagnoses of LRTIs, such as pneumonia and bronchitis (Table 1). The subject was recorded as having had an LRTI during a 1-year period if he/she had had at least 1 diagnosis during that year. This was recorded for every year from birth to age 6 years, and then the sum of the periods with at least 1 infection was calculated for the LRTIs, so that the maximum number of time periods with exposure was 6.


The following variables were included in the analyses as potential confounders: age,[24] sex,[24] family socioeconomic status at baseline,[25] duration of breastfeeding,[26] secondhand smoke exposure from birth to age 3 years,[27] maternal smoking during pregnancy,[27] history of pets at home from birth to age 3 years,[28] and form of day care (day-care center or family day care vs. home care).[6] Information on the covariates was obtained from the baseline questionnaire and later follow-up surveys. Family socioeconomic status was determined by combining the highest level of parental education with the highest parental occupational level at baseline.

Statistical Methods

We assessed the relationships between early respiratory tract infections and risk of developing asthma later in life by comparing the risk of asthma among subjects with LRTIs and/or more than the cutoff point of URTIs (exposure categories) with the risk among subjects with no LRTIs and fewer than the cutoff point of URTIs (reference category), taking into account the age of the child. The analysis was based on the time to onset of asthma, and the ratio of hazard rates—denoted the hazard ratio—was used as the measure of association.[29] First, in part 1, we applied 4 overlapping age periods (baseline age to ages 6, 12, 18, and 27 years) to ascertain the influence of respiratory infections on the risk of asthma over time. The trend in the hazard ratio over time (from baseline age to age 27 years) was assessed using the Wald χ2 test. Second, we evaluated whether the age at which respiratory infections were experienced had any association with the studied hazard ratios by comparing the adjusted hazard ratios for children who had experienced respiratory infections early (1–2 years of age) with those of children who had experienced respiratory infections later (3–7 years of age). We tested the statistical significance of the difference between the compared hazard ratios using Q statistics for heterogeneity between the corresponding adjusted hazard ratios.[30,31] In part 2, we applied the corresponding age periods used in part 1, with the exception that instead of using baseline age, we used the subject's date of birth.

The hazard ratio for development of asthma was estimated using Cox's proportional hazards model and adjusting for the covariates mentioned above. The proportionality of all variables was evaluated visually from the survival curves and quantitatively by applying proportionality test measures. The hazard ratios were calculated using the PROC PHREG procedure in SAS, version 9.3 (SAS Institute, Inc., Cary, North Carolina).