The Asthma Attack Study

A prospective multicentre study is being done in children aged 5–17 years in low-income settings in three Ecuadorian cities (Quito, Cuenca and Portoviejo) to study factors associated with acute asthma attacks. Because of the impact of the COVID-19 pandemic on risk of asthma attacks and healthcare access, the study was divided into two phases of recruitment and follow-up: (1) phase I (or pre-COVID-19 from May 2019 to March 2020) evaluated risk factors associated with recurrence of acute asthma attacks among children and adolescents attending ERs of public hospitals in low-income settings in the three cities and (2) phase II (or peri-COVID-19 from March 2021 to June 2022) is evaluating risk factors associated with attacks among children and adolescents with asthma and current symptoms who have previously attended specialist care at public hospitals in Cuenca and Portoviejo. Phase II is also evaluating the effects of exposures to SARS-CoV-2 on asthma attacks and symptoms. In addition, both study phases are evaluating the impact of asthma symptoms and adequate control of asthma on the quality of life and economic costs for the family of children and adolescents with asthma. The objectives of phases I and II are detailed in table 1. Qualitative studies of health worker’s perspectives on asthma care coordination between primary and specialised healthcare are being done also as described elsewhere.41

Study design and setting

This is a multicentre prospective study conducted in three Ecuadorian cities (Quito, Cuenca and Portoviejo). Quito, the capital, and Cuenca, the third largest city, are located in the Andean highlands (altitude>2500 m) where the annual average temperature is ~16°C, while Portoviejo is in the coastal region (at sea level) with an annual average temperature of ~25°C.42 According to the 2010 census, the populations of the three cities are: Quito>2 million, Cuenca 0.5 million and Portoviejo 0.28 million inhabitants. Ecuador is an upper-middle-income country with a per capita income of US$6080 in 2019. The health system in Ecuador includes public and private provision. Public institutions offer universal health coverage that is stratified into different levels of care from primary to tertiary services.43 Social security institutions offer health services to affiliated salaried workers and their families. The private sector includes for-profit entities (hospitals, clinics, dispensaries, doctor’s offices, pharmacies and prepaid health insurance companies), which are generally located in the main cities.44

Study centers

The study was initially based in tertiary care hospitals in each of the three cities (Hospital Docente General de Calderon, Quito; Hospital Vicente Corral Moscoso, Cuenca; and Hospital Verdi Cevallos Balda, Portoviejo).

Recruitment: phase I (pre-COVID-19; May 2019 to March 2020)

Children and adolescents aged 5–17 years were recruited either while attending ERs with an asthma attack or through daily registries and contact telephone numbers of patients attending the ERs with an asthma attack. From May 2019, eligible patients (table 2) were provided with a study information sheet on discharge and their parents were contacted to arrange an interview at a hospital consulting room 10–14 days later when they were recruited into the study after providing informed written parental consent and minor assent (8 years and older).

Recruitment: phase II (peri-COVID-19; March 2021 to March 2022)

The COVID-19 pandemic had a dramatic impact on our ability to recruit children with asthma attacks in the three study hospitals that resulted in the suspension of recruitment in March 2020. This was owing to the reassignment of all three hospitals to COVID-19-only activities for prolonged periods and the imposition of a legally enforced lockdown nationally. Figure 2 shows the number of potentially eligible children identified in the three study hospitals by month in the period prior to COVID-19 and during the pandemic. There was a marked reduction in eligible children attending the ERs immediately following the imposition of lockdown on 16 March 2020 and during the following 8 months. In the period following the initial lockdown and other mitigation strategies to reduce transmission of SARS-CoV-2, within the sample of asthmatic children recruited into phase I of the study, there was evidence of dramatic reduction in health service attendance rather than changes in the incidence of attacks.19 To allow continued recruitment, a second phase (phase II) of the study was started with eligibility criteria modified to include children and adolescents aged 5–17 years registered as having asthma in patient registries at the study hospitals in Cuenca and Portoviejo (Quito was not included in phase II because of logistic issues in that specific setting). The child’s parent was contacted by telephone and asked about the child’s current asthma symptoms. Children with current symptoms were considered eligible. Eligibility criteria for phases I and II are shown in table 2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Study procedures during baseline evaluation and follow-up in phases I (A) and II (B) of study. Follow-up in phase I was for 12 months and in phase II for a minimum of 6 months. ACT, Asthma Control Questionnaire (>11 years); AFCQ, Asthma Family Costs Questionnaire; ARQ, Asthma Recurrence Questionnaire; BaselineQ, general questionnaire based on phase II of the International Study of Asthma and Allergies in Childhood; C-ACT, Child Asthma Control Questionnaire (≥12 years); FeNO, fractional exhaled nitric oxide; NAKQ, Newcastle Asthma Knowledge Questionnaire.

Data collection

A study manager supervised day-to-day activities of dedicated teams (physician and nurse) of trained study personnel based in each of the public hospitals. Each participant was followed up for 12 months in phase I and at least 6 months in phase II. Standardised procedures were used throughout (figure 2). Subjects were evaluated at monthly intervals using face-to-face and telemonitoring for follow-up. Both phases include face-to-face evaluations (evaluations at baseline, 2, 4, 6, 9 and 12 months in phase I; baseline, 6 and 12 months in phase II). Between March 2020 and March 2021, all face-to-face evaluations in phase I were changed to telemonitoring apart from the final evaluation at 12 months, which remained face-to-face, where possible. Final face-to-face evaluations were done in March–April 2021 on participants who had completed 12 months of follow-up between June 2019 and February 2020. The site of face-to-face evaluations was changed to either a clinic or home visit depending on the preference of the child’s parents. Study procedures were the same in both phases of the study (figure 2) with the following exceptions: (1) procedures requiring face-to-face evaluations at 6 and 12 months in phase I (Childhood Asthma Control Test (C-ACT) and Pediatric Asthma Quality of Life Questionnaire (PAQLQ)) were dropped in phase II and the remaining instruments were administered by telemonitoring; (2) blood samples (baseline, 6 and 12 months) are being collected from those recruited into phase II of the study for measurement of anti-SARS-CoV-2 IgG antibodies; and (3) measurement of lung function and fractional exhaled nitric oxide (FeNO) was suspended from March 2020 because of risks from potentially hazardous infectious aerosols until a time when this procedure can be done safely. Medications (inhaled salbutamol and fluticasone) were provided free of charge to participants when prescribed by a hospital physician and there was no medication in the hospital pharmacy. The study team ensured adequate inhaler technique during face-to-face clinic or home visits. Study personnel did not interfere with treatment indications of hospital staff.

Questionnaires at baseline and follow-up

The following questionnaires were used as indicated in figure 2; (1) asthma questionnaire to collect information on history of asthma symptoms, treatment and management, and potential risk factors for asthma attacks. This questionnaire has been previously adapted from phase II of the International Study of Asthma and Allergies in Children, and extensively field tested for children and adolescents with asthma13 26; (2) Asthma Recurrence Questionnaire—questionnaire to monitor recurrence of asthma symptoms and monitor treatment during follow-up,13 including symptoms and diagnosis of COVID-19 in phase II; (3) C-ACT,45 completed by the child and parent for children aged<12 years or the ACT46 for those aged≥12 years; (4) PAQLQ47; (5) Newcastle Asthma Knowledge Questionnaire48 49 validated in a Spanish-speaking population; and (4) Asthma Family Cost Questionnaire.50 Data collection is summarised in table 3.

Lung function test and FeNO

Lung function testing is being done according to American Thoracic Society criteria using a portable ultrasonic spirometer (EasyOne device, ndd Medical Technologies). Global Lung Function Initiative 2012 reference values will be used to calculate z scores depending on age, sex, height and ethnicity.51 Lung function is being measured before and after 200 µg salbutamol administered through a spacer with a positive test for airways reactivity defined as an increase in forced expiratory volume in one second of ≥12%. FeNO is being measured in parts per billion using NObreath (Bedfont Scientific, UK). As a consequence of lockdown and mitigation strategies to prevent transmission of SARS-CoV-2, measurement of lung function and FeNO was suspended in March 2020. FeNO measurement, a low effort procedure that carries a minimal risk of generating aerosols,52 53 was resumed in March 2021 for 12-month evaluations in phase I and baseline evaluations in phase II. FeNO is being done following the recommendations of the manufacturer to minimise potential risks.54 Spirometry, which often induces cough, will be resumed, only if the procedure can be done without risk to staff, patients and their families.

Blood samples

Blood sample are being taken at baseline, 6 and 12 months in phase II. Blood is drawn from an antecubital vein into a 4 mL plastic tube containing EDTA (ethylenediaminetetraacetic acid) as anticoagulant (Vacutainer, BD). After centrifugation, plasma is stored at −20°C until analysis. Levels of IgG specific for SARS-CoV-2 are being measured using a validated enzyme linked immunosorbent assay.55

Biosafety precautions relating to the COVID-19 pandemic

Personnel gathering data during face-to-face visits have been trained to use appropriate personal protective equipment according to international guidelines.56–58 Immediately prior to a home or clinic visit, the child’s carer is telephoned about potential symptoms or a COVID-19 diagnosis among family members or close contacts—if any symptoms are reported the visit is rescheduled after 2 weeks.

Sample size considerations

A sample size of 450 patients was proposed for phase I, based on a binary outcome variable of any recurrence of an asthma attack within 12 months or so. We used an estimated recurrence risk of about 50% during 6-month follow-up from a similar population in Ecuador.13 The study should collect between 10 and 16 outcome events per each potential explanatory variable for robust results (ie, for at least 80% power, equivalent to a type 2 error of less than 20% as proven by simulation studies). Assuming 10 explanatory variables are considered, in the worst-case scenario, 10×16=160 such events would be required for the upper limit of 16 events for each explanatory variable. Given a 50% risk of recurrence, a minimum of 160×2=320 subjects will be needed. Allowing for 20% attrition/loss during follow-up given the longitudinal nature of the study, we needed approximately 400 patients, which we set more conservatively at 450.59–62 However, we do not expect a multivariable model to exceed 10 predictors. There is the possibility of differences between populations recruited during the two phases of the study. If this is not the case, an indicator variable which annotates the 2 phases will be added to the set of predictors and 450 patients would still allow at least 80% power under the conditions described. For example, if 11 predictors are considered, 352 patients would be required or a total of 440 allowing for 20% attrition. If there are substantial differences between subjects recruited in the 2 phases, analyses will consider up to 10 events per explanatory variable requiring 100 events per phase among 200 participants followed up or 250 recruited (allowing for 20% attrition). The latter scenario will require recruitment of a total of 500 during the two phases, a target that we are confident we can achieve.

The definition of an asthma attack in both phases of the study is an episode of acute wheezing requiring an unscheduled visit to a health professional (including visits to ERs) and which is controlled by β₂-agonist bronchodilator treatment. The original definition for an asthma attack in phase I that included ER attendance had to be modified in the light of changes in health-seeking behaviour caused by the COVID-19 pandemic and fear of contagion in hospitals and other acute health facilities.19 Potential explanatory variables include age, gender, ethnicity, socioeconomic status, medication use, asthma attacks and use of systemic corticosteroids in year prior to recruitment, asthma control test score and study centre. Given changes to the phase I protocol, an additional variable (for recruitment from ERs vs hospital lists) will be considered as an effect modifier. Subgroup analyses of phase II will include SARS-CoV-2 seropositivity as an explanatory variable. Secondary outcomes will be asthma control score and economic costs to the patient’s family.

Statistical analyses

The statistical models to be used by study objective are summarised in table 1. The primary outcome will be any asthma attack recurrence during 12 months of follow-up in phase I and at least 6 months in phase II. Logistic and linear regression and survival analyses will be used. For the latter, if no reattendance related to an asthma attack occurred during the follow-up period, the patient is considered (right) censored. The general assumption for time to event (survival analysis) is that the censoring is non-informative, that is, it is independent of the events (asthma attacks). Under this assumption, the results should not be qualitatively different from those inferred from a logistic regression. ORs and/or HRs and their corresponding p values and 95% CIs will measure the associations between asthma attack recurrence and explanatory variables, and their statistical significance and uncertainty. The Cox proportional hazard semiparametric technique allows no assumption to be made on the baseline hazard; nevertheless, proportionality of the hazards should hold. The latter assumption can be tested on the basis of Shoenfeld residuals after fitting models to the data. If unsatisfactory, alternative parametric models will be considered after understanding the distributional assumptions on the data such that the baseline hazard function is modelled appropriately. After exploring the data as described above, a forward–backward elimination survival analysis would be applied to determine the most parsimonious model for an adjusted multivariable model; that is, the model with the least number of predictors yet explaining most of the variability in the data. Penalised-likelihood criteria such as Akaike information criterion or Bayesian information criterion will be used to choose between non-nested models—nevertheless based on complete observations only.

Hosmer-Lemeshow measure of goodness of fit would be employed for logistic regression models with a p value of less than 0.05 indicating a poor fit to the data. Model validation will be done by splitting the data set into training–testing subsets (logistic regression setting). Cross validation techniques will be used to understand the prediction power of the corresponding logistic regression model using area under the receiver operator curve, prior to propose the model entering clinical practice with the view of elaborating a prognostic model. In a survival setting, it is also possible that the precise time of the attacks may be missed but recorded as between two time points. This would fall into interval censored data as an alternative to survival setting in which the precise time of the event is required.

Multiple recurrences of asthma attacks within the follow-up period will generate a longitudinal binary outcome, with multiple records for each participant over follow-up. The hierarchical structure of the data requires special inferential techniques, which account for the two sources of variability: that of between participants and within participants. Three different modelling approaches will be considered: generalised estimation equation, mixed models and random effects survival models. The first approach provides population average estimates, the second provides subject-specific estimates, while the third is also known as shared frailty survival model. A frailty is a latent random effect, which enters multiplicatively on the hazard function and accounts for the multiple failures setting and/or for the clustered aspect of the data.

Complete data analyses will be done under missing completely at random assumptions. However, patterns in missing data will be explored and methods to evaluate potential impact of missing data on estimates will be used as appropriate such as multiple imputations using chained equations accompanied by a sensitivity analysis to the missing data assumptions.63–65 Analyses will use STATA (StataCorp. 2017. Stata Statistical Software: Release V.15. College Station, Texas: StataCorp) or R (R Core Team (2013)).