The response to combination therapy treatment regimens in severe/difficult-to-treat asthma

Asthma treatment guidelines from the National Asthma Education and Prevention Program 1 and the Global Initiative for Asthma 2 and recommend inhaled corticosteroids (ICS) plus a long-acting β-agonist (LABA) as the preferred treatment regimen in patients with moderate or severe persistent asthma who are not adequately controlled on ICS therapy alone. These recommendations were based on several large pivotal randomised clinical trials that demonstrated, through the primary efficacy outcome of pulmonary function improvements, the superiority of adding LABA to ICS for the management of mild-to-moderate persistent asthma 1, 2. Salmeterol/fluticasone combination (SFC), an aerosol containing a LABA (salmeterol) and an ICS (fluticasone), is indicated for the maintenance treatment of asthma in patients with reversible obstructive airway disease when the use of a combination product is considered to be appropriate 3. Studies have demonstrated that low-to-intermediate doses of SFC (100/50 or 250/50 μg b.i.d) confer greater clinical benefit for mild-to-moderate asthma when compared with increased doses of an ICS or the addition of a leukotriene modifier to an ICS 3, 4.

There is a lack of evidence that demonstrates the superiority of high-dose SFC (500/50 μg b.i.d) compared with its alternatives (such as high-dose ICS therapy or high-dose ICS plus other non-LABA long-term controller therapies) in adults with severe or difficult-to-treat asthma, patients who bear the greatest burden of the disease. In a randomised trial comparing high-dose SFC (500/50 μg b.i.d) to both the noncombined form of fluticasone (500 μg b.i.d) and salmeterol (50 μg b.i.d) and fluticasone (500 μg b.i.d) alone, Aubier et al. 5 reported improved peak expiratory flow rates for the high-dose SFC group compared with the fluticasone alone group, but no statistically significant differences between treatment groups with respect to forced expiratory volume in one second (FEV₁), daytime or night-time symptoms, and the need for rescue medication. Although the Gaining Optimal Asthma ControL (GOAL) study demonstrated that more uncontrolled patients were able to achieve control with SFC compared with fluticasone propionate alone, GOAL also demonstrated that most patients who achieved control did so at the lowest dose of combination treatment used for the stratum 6.

However, current SFC prescribing patterns, even in the absence of established clinical evidence, show that care providers prescribe high-dose SFC to many difficult-to-treat patients or to those who have failed to respond to low-dose SFC. It is also well established that pulmonary function outcomes on which the SFC efficacy claim is primarily based do not fully reflect a patient's level of asthma control 1, 2, 7–10. Thus, research is merited to investigate the effects of low- and high-dose SFC on the many dimensions of a patient's asthma control. The present study assessed the response to low-dose SFC and high-dose SFC in severe or difficult-to-treat asthma by longitudinally comparing patients who were on either low- or high-dose SFC to those who were on long-term asthma medications such as ICS or LABA, but not on both, using health outcomes consistent with components of asthma control.

MATERIALS AND METHODS

RESULTS

Of the 208 patients in the control group at baseline, 177 (85.1%) were receiving ICS while 10 (4.8%) were receiving LABA. Of the control group patients neither on ICS nor LABA medications, seven (3.4%) were on other long-term asthma controllers, including leukotriene modifiers, while the remaining 14 (6.7%) were not on any long-term controller medications at baseline. Baseline demographic and clinical characteristics, including measures of the study outcomes at baseline for the three treatment groups, are displayed in table 1⇓. Smoking status, age, and sex were similar between treatment groups. The high-dose SFC group had a younger age at asthma onset and a higher proportion of minorities compared with the never-on-SFC group, conditions predictive of fixed airway obstruction 16. Immunoglobulin E levels, skin test results, history of allergic rhinitis or atopic dermatitis, education, employment status or type of healthcare insurance did not significantly differ between treatment groups (data not shown; p-value >0.05 for all comparisons). Statistically significant differences in many of the outcome measures at baseline were observed between the treatment groups. Summary statistics for outcome measures at baseline showed that the high-dose SFC group had more poorly controlled asthma (i.e. lower MiniAQLQ scores, lower FEV₁ scores, higher ATAQ scores, higher proportion of patients with an asthma exacerbation, and a higher proportion considered to have severe asthma) than the never-on-SFC group; the never-on-SFC group had more poorly controlled asthma than the low-dose SFC group.

The propensity score models showed fair discrimination between the high-dose SFC and never-on-SFC groups (area under the ROC curve 0.68) and between the low-dose SFC and never-on-SFC groups (area under the ROC curve = 0.66). Hosmer–Lemeshow Chi-squared tests indicated that both models were well calibrated (lack of fit: p = 0.30 for low-dose SFC versus never-on-SFC, p = 0.40 for high-dose SFC versus never-on-SFC). Baseline characteristics were similar between treatment groups after stratifying by quintile.

Figures 1⇓–⇓⇓⇓5⇓ display the unadjusted and adjusted differences in the 24-month outcomes between low-dose/high-dose SFC and the never-on-SFC group. Compared with the never-on-SFC group, unadjusted mean differences in the overall MiniAQLQ scores were 0.40 (95% confidence interval (CI) 0.15–0.64) for the low-dose SFC group and -0.23 (95% CI -0.47–0.02) for the high-dose SFC group (fig. 1⇓). The adjusted mean difference in the overall MiniAQLQ scores was 0.20 (95% CI 0.00–0.40) for the low-dose SFC group compared with the never-on-SFC group, but the difference between the high-dose SFC group and the never-on-SFC group was -0.01 (95% CI -0.21–0.18).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1—

Least mean square difference and 95% confidence intervals (low-dose salmeterol/fluticasone combination (SFC) minus never-on-SFC; high-dose SFC minus never-on-SFC) in mean Mini Asthma Quality of Life Questionnaire overall scores at 24 months. ^#: baseline differences between treatment groups, including severity differences, were adjusted with propensity scores. ^¶: quality of life (QoL) 0.40, p = 0.0015; ⁺: QoL 0.20, p = 0.0456; ^§: QoL 0.23, p = 0.0672; ^ƒ: QoL 0.01, p = 0.8915.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2—

Least mean square difference and 95% confidence intervals (low-dose salmeterol/fluticasone combination (SFC) minus never-on-SFC; high-dose SFC minus never-on-SFC) in mean Asthma Therapy Assessment Questionnaire scores at 24 months. ^#: baseline differences between treatment groups, including severity differences, were adjusted with propensity scores. ^¶: asthma control −0.46, p<0.0001; ⁺: asthma control −0.33, p = 0.0018; ^§: asthma control 0.18, p = 0.1328; ^ƒ: asthma control −0.04, p = 0.7428.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3—

Least mean square difference and 95% confidence intervals (low-dose salmeterol/fluticasone combination (SFC) minus never-on-SFC; high-dose SFC minus never-on-SFC) in mean forced expiratory volume % predicted pre-bronchodilator values at 24 months. ^#: baseline differences between treatment groups, including severity differences, were adjusted with propensity scores. ^¶: lung function 4.61, p = 0.0629; ⁺: lung function 1.93, p = 0.1866; ^§: lung function −5.93, p = 0.0077; ^ƒ: lung function 2.93, p = 0.0462.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4—

Odds ratio (OR) and 95% confidence intervals (high-dose salmeterol/fluticasone combination (SFC) versus never-on-SFC; low-dose SFC versus never-on-SFC) of severe asthma risk at 24 months. ^#: baseline differences between treatment groups, including severity differences, were adjusted with propensity scores. ^¶: OR 0.62, p = 0.0367; ⁺: OR 0.76, p = 0.3441; ^§: OR 2.99, p<0.0001; ^ƒ: OR 2.16, p = 0.0013.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5—

Odds ratio (OR) and 95% confidence intervals (high-dose salmeterol/fluticasone combination (SFC) versus never-on-SFC; low-dose SFC versus never-on-SFC) of the risk of experiencing asthma exacerbations at 24 months. ^#: baseline differences between treatment groups, including severity differences, were adjusted with propensity scores. ^¶: OR 0.85, p = 0.4603; ⁺: OR 0.99, p = 0.9592; ^§: OR 1.99, p = 0.0005; ^ƒ: OR 1.42, p = 0.1096.

After adjustment, the low-dose SFC group compared with the never-on-SFC group had an ATAQ score mean difference of -0.33 (95% CI -0.54– -0.12), whereas the adjusted mean difference for the high-dose SFC group compared with the never-on-SFC group was -0.04 (95% CI -0.24–0.17; fig. 2⇑). Mean difference in adjusted FEV₁ % pred pre-bronchodilator values was 1.93% (95% CI -0.93–4.80%) for the low-dose SFC group versus the never-on-SFC group, and was 2.93% (95% CI 0.06–5.81%) for the high-dose SFC group versus the never-on-SFC group (fig. 3⇑).

After adjustment, the odds of having severe asthma was 0.76 (95% CI 0.43–1.34) for the low-dose SFC group compared with the never-on-SFC group, but was 2.16 (95% CI 1.35–3.45) for the high-dose SFC group compared with the never-on-SFC group (fig. 4⇑). After adjustment, no statistically significant differences in exacerbations were observed between either the high-dose or low-dose SFC group and the never-on-SFC group (p = 0.959 and p = 0.110, respectively; fig. 5⇑).

DISCUSSION

The present analysis examined the response of low- and high-dose SFC on health outcomes consistent with components of asthma control in a large cohort of patients with severe or difficult-to-treat asthma. Observed differences in asthma-related health outcomes at baseline confirmed that low-dose SFC was prescribed to patients who had better asthma control than the never-on-SFC group, and that high-dose SFC was prescribed to patients who had the lowest levels of asthma control. Unadjusted observed differences at the 24-month follow-up visit followed this trend; the low-dose SFC cohort achieved better asthma control than the never-on-SFC group, and the never-on-SFC group achieved better asthma control compared with the high-dose SFC group. The propensity score method was used to adjust for treatment selection bias and other potential confounding.

After adjustment for the propensity to be in a particular treatment cohort and for the baseline outcome values, the low-dose SFC group had significantly more favourable 24-month outcomes than the never-on-SFC group with respect to the differences in MiniAQLQ and ATAQ, and trended toward more favourable FEV₁ and asthma severity outcomes. Many of the adjusted outcomes were comparable between the high-dose SFC and the never-on-SFC group. These findings suggest that some, if not many, of the high-dose SFC patients showed limited clinical benefit with high-dose SFC. The explanation for the failure of the high-dose SFC cohort to demonstrate substantial improvements in measures of asthma control above and beyond patients not treated with SFC may include inherent, not easily quantifiable, differences between the patients who receive and do not receive high-dose SFC in clinical practice. However, an alternative explanation is that the benefits derived from SFC are optimally achieved with the low-dose SFC and that incremental doses of SFC do not provide added clinical benefit in the severe or difficult-to-treat asthma population.

The current findings suggest a lack of dose-response for SFC. A failure to respond to high-dose SFC in the severe or difficult-to-treat population could be explained by the presence of fixed airway resistance, molecular mechanisms of steroid resistance 17, 18, a steroid-insensitive form of inflammation 19 or distal lung inflammation not accessible to inhaled medications 20. The modest high-dose SFC benefit is consistent with the findings of: Aubier et al. 5, who reported limited benefits of high-dose SFC compared with its alternatives; Szefler et al. 21, who found a lack of dose-response for ICS; and Martin et al. 22, who demonstrated that ICS nonresponders did not achieve better asthma control when compared with placebo nonresponders. It is possible that more nonresponders were given high-dose SFC compared with low-dose SFC and, therefore, the incremental benefit of high-dose SFC was not observed because of nonresponders. The validity of the observational research will hinge on the findings of similar but independent studies, with or without randomised designs.

The clinical interpretation of the observed differences should be considered carefully. For example, the largest observed mean difference in the total MiniAQLQ scores was 0.40 (in favour of the unadjusted low-dose SFC compared with the never-on-SFC group). Since 0.5 has been found to be the minimal important difference in the AQLQ outcome 23, it could be argued that both the unadjusted and adjusted observed differences in the analogous MiniAQLQ outcome did not achieve clinical significance.

There are several limitations to the present study that are noteworthy. Medication utilisation was captured in TENOR from an annual study coordinator interview. Therefore, limited data are available to assess or verify compliance or persistence. However, the current study design restricted the SFC groups to those who started SFC treatment between the baseline and 12-month follow-up visit, and patients were required to remain on this dose of SFC treatment throughout the 24-month follow-up visit. The comparison group comprised TENOR adults who were not receiving SFC at any of the visits (baseline, 12- or 24-month follow-up) but could have been receiving an ICS or LABA medication. As with any observational study, confounding variables may mask the true effect that SFC has on asthma outcomes.

While confounding may still remain, propensity score methods were applied to adjust for potential confounders that were measured within the TENOR dataset. Arguably, covariate adjustment remains a slightly more common technique for addressing potential confounding, but Cook and Goldman 24 showed that the propensity score methodology is often more robust. The propensity score approach received further endorsement by Lauer 25 who suggested that propensity scores may be the only viable method to answer many questions that arise from clinical practice. Although the present authors evaluated the fit of the propensity score models and the ability of the propensity score models to discriminate between the treatment cohorts, there is no way to test whether other unmeasured variables could have led to a different and better propensity score adjustment. Additionally, the propensity score adjustment depends upon the accurate and precise measurement of the variables that contribute to the model; thus, possible residual confounding in the current analysis cannot be ruled out.

The current results support evidence that some people with asthma achieve better control while taking low-dose salmeterol/fluticasone combination, but also suggest that high-dose salmeterol/fluticasone combination fails to provide a clinically significant benefit in terms of many control-related health outcomes as compared with similar patients not taking salmeterol/fluticasone combination. These findings suggest a limited value of high-dose salmeterol/fluticasone combination. While additional studies are needed, the present findings call for alternative therapeutic approaches in severe/difficult-to-treat asthma for those unable to attain asthma control with or without salmeterol/fluticasone combination.

Statement of interest

Statements of interests for all of the authors and the study itself can be found at www.erj.ersjournals.com/misc/statements.shtml