Effectiveness of Montelukast on asthma control in infants: methodology of a French claims data study

This pilot study was performed using the EGB database, a 1/97 sample of the whole French population, to prepare a project aimed at assessing the effectiveness of MTL-4 in infants suffering from asthma. The first three questions addressed by this pilot study were answered satisfactorily, i.e. identification of infants with uncontrolled asthma (N = 1,149), frequency of exposure to MTL-4 and ICS (n = 51 and 768, respectively), and occurrence of outcomes (78% and 79% of MTL-4 and ICS infants, respectively). For the last question, i.e. development of matching criteria, we defined a new marker of severity (therapeutic typologies).

Asthma in infants under 36 months is usually defined as an episode of wheezing dyspnea that has occurred at least 3 times since birth [2]. In our study, as diagnoses are usually missing in French claims data, the identification of asthmatic infants relied on a proxy, i.e. the use of specific therapy (2 consecutive dispensations of respiratory drugs ATC R03, separated by an interval of 8 to 91 days). Based on this approach, 1,790 asthmatic infants were identified between March 2010 and March 2012. Combined with the mean annual number of births in France – around 790,000 –, the prevalence of asthma may be estimated to be around 11% in infants. Epidemiological data on the prevalence of asthma in infants are rare. One report suggests that the prevalence of these symptoms at the age of two lies between 11.9% and 26.6% [16], which is compatible with our results. The sex ratio observed in our study is consistent with data available for young children under 5 years, where a 2/1 ratio is traditionally observed in favor of boys [17]. In summary, prevalence and sex-ratio support the validity of the algorithm used for the identification of infants with asthma in our study.

Using our selection criteria, 51 users of MTL-4 were identified in the EGB, together with 768 users of ICS. These represent adequate counts to meet the objectives of the final study, in particular for group comparison. Sales data suggest that a maximum of 78,000 infants aged 6 to 24 months were exposed to MTL-4 over a 3-year period (2010 to 2012), which corresponds to around 250 infants/year in the EGB, in line with our figures stemming from strict inclusion criteria [18]. Similarly, the identification of outcomes in the EGB and their frequency also confirmed that the study was feasible using claims data (≥49% of infants in the MTL-4 and ≥57.9% of the ICS groups were identified with uncontrolled asthma, or dispensing of oral corticosteroids). According to the IRDES (Institut de Recherche et Documentation en Economie de la Santé, the national health economics research institute) survey [17], in 60% of patients, asthma is inadequately controlled (based on GINA criteria). According to stricter criteria from the French health authority (Haute Autorité de Santé, HAS) guidelines [2], asthma is inadequately controlled in 83% of patients. In the Er’Asthme study [19,20] that investigated children aged 6 to 14 years, considering the Canadian criteria which were similar to HAS criteria, asthma was inadequately controlled in 73% of children. The results obtained in our explanatory study are not incompatible with these references, again supporting their validity.

This pilot study was conducted using the EGB database [14], which is a sample of French national claims data. In several countries, comparative effectiveness research studies have been conducted in asthma using claims data [21,22]. In France, our study was among the firsts with this objective, justifying a feasibility study. A major advantage of EGB is to record exhaustive medical resource utilization in a 1/97th sample of the 86% of French population covered by the national health insurance system. In addition to detailed data on drug utilization, the EGB also contains data on all medical procedures, and all medical and paramedical visits. Information on prescribers is also available. The linkage of primary care with hospital data (PMSI) provides access to data on hospitalizations (private and public hospitals), i.e. exhaustive individual medical resource utilization. As the SNIIR-AM records health care provided to infants following exposure to MTL-4, it was a priori considered to be a useful source of data to test the effectiveness of this drug. Another approach would have been a field study in a population of asthmatic infants. Besides practical pitfalls, such as complex organization and long timelines, a field study would have been exposed to recall bias, particularly on past medical resource utilization.

Nonetheless, the EGB database has certain intrinsic limitations. Few variables were available to describe patients: age, gender, free-access-to-care status, and LTD status in case of more severe disease. EGB does not include clinical data (e.g. lab test results), or information on family history. Prescriptions without dispensations or dispensations without drug utilization by the patient are not recorded, neither are prescribed doses or durations of prescriptions [23].

This study had other limitations. Uncontrolled asthmatic infants were identified using ≥2 distinct R03 drug dispensations followed by exacerbation and/or acute loss of control. These criteria excluded a significant proportion of MTL-4 -exposed infants. This also limited the generalizability of our findings. For example, the rate of outcomes during the 6 months of follow-up could have been overestimated by the choice of the selection criteria, as asthma had to be uncontrolled for infants to be included. Also, defining exacerbations from dispensations of oral corticosteroids could be criticized as such use could be due to conditions other than asthma. However, it is likely that requiring the initial dispensation of oral corticosteroids to be followed within seven days by a dispensation of MTL-4 or ICS ensured adequate specificity of this marker to identify asthma exacerbations.

Finally, the matching for asthma severity was based on proxies (therapeutic typologies and past asthma-related hospitalization) and probably did not fully account for potential differences in severity between the two groups. Therapeutic typologies allowed us to classify infants into three groups of severity based on treatments before index date: infants with no asthma treatment, infants with only reliever therapy, and infants already receiving controller therapy. We believe that this classification optimized the use of information available in French claims data.

Importantly, the study design was based on the assumption that uncontrolled asthma leads to increased medical resource utilization, and this concept is now generally accepted, as exemplified by recent definitions [2,24]. Other methodological biases are, however, possible. As in any observational study comparing effectiveness between two therapeutic strategies, indication bias can never be excluded [25]; nonetheless, as already stressed, matching of infants on the typology of initial therapy limits this bias.