Results
Baseline characteristics and variable distribution in the patient sample, overall and according to montelukast use, are reported in Table 1.
We first analysed possible relationships between demographic features and montelukast usage. A significant difference was observed in the age of patients exposed (median = 60, 46 Q1 - 68 Q3) and non-exposed to montelukast (median = 64, 55.75 Q1 - 70 Q3). In addition, 77.7% of rural zones asthmatic (108/139) were exposed to montelukast, compared to only 44.2% of urban patients (292/661), while within low-education patients only 28.1% were exposed to montelukast (68/242), compared to 59.5% within mid+high education (332/558). Income did not show a significant relationship to montelukast use. The use of additional drugs appeared to be very balanced between the two cohorts, exposed and non-exposed to montelukast, with the exception of antiplatelet (5% vs. 9.75%) and antitumoral drugs. Overall, 37 (4.6%) of our asthmatic patients suffered a major CV event during the observation period. Interestingly, the overall incidence of an ischemic event in our asthmatic population (15.4 events per 1000 patient-years) is similar to the expected rate in a general population (between 8 and 18.1, depending on level of blood pressure32). However, considering only patients not exposed to montelukast, the incidence is higher (26.7 events per 1000 patient-years) than that observed in the non-asthmatic population, while the incidence in patients exposed to the drug is 4.17 events per 1000 patient-years, which is lower than expected in the general population32. The analysis of the incidence of CV events showed that the age of patients that did not experience a CV event (median = 62) was significantly (Wilcoxon test, p < 0.00004) lower when compared to patients suffering a CV event (median = 74), with 12 years difference between the two groups (Figure 1).
Results from Cox model using a PS matching, are shown in Tables 2. We performed PS matching with different calipers but we focused on a caliper = 0.2 because it has been demonstrated to be optimal in many settings33 to look for predictors of a major CV event. PS matching lead to 269 patients in each group with a number of events of 22 resulting in a significant (p = 0.0065) protection in CV events (HR = 0.222) due to montelukast use.
Because in the PS matching analysis we lost some of the events (15 out of 37, Table 2) we also used a Cox regression model adjusted for PS, which allowed us to take all the events into consideration (Table 3). Even though patients exposed to montelukast were overall younger than the ones not exposed to the drug, montelukast remained a significant (p = 0.0046 protective factor (HR = 0.241) for CV events. In addition, as the use of antiplatelet drugs was significantly different between exposed and not exposed to montelukast (Table 1), we also performed a Cox model adjusted for PS only on patients taking antiplatelet drugs. As shown in Table 3, the results are still statistically significant and not substantially different from those considering the whole sample.
In Figure 2 are reported the event-free Kaplan-Meier survival curves of patients exposed or not exposed to montelukast. On the left panel, the whole patient sample was considered, while on the right panel were considered only patients using antiplatelet drugs. In both cases the two curves resulted statistically different, while, as expected, the difference in survival probability was higher when limiting the analysis to patients treated with antiplatelets (Logrank test, p = <0.0001 vs. p = <0.0001).