Influences of earlier adherence and symptoms on current symptoms: a marginal structural models analysis

The morbidity and mortality associated with asthma are suspected to be a result, in part, of poor adherence to inhaled corticosteroid regimens. One influence on adherence may be the perception of symptoms. Because symptoms and adherence affect each other over time, a conventional statistical approach for studying these relationships may provide biased results.

To understand the influence of previous asthma symptoms and previous adherence on current symptoms.

A total of 76 adults, mean age 48 years +/- 15 years, with moderate or severe persistent asthma underwent 6 weeks of electronic monitoring of their use of inhaled corticosteroids and completed a daily symptom diary. We estimated the effect of earlier adherence on final symptoms by using marginal structural models, estimated by using a weighted estimation technique.

Morning was better than evening adherence, which declined over the observation period. The variability of adherence appeared to increase over the observation period. In addition, earlier adherence predicted current adherence more strongly than earlier symptoms predicted current adherence. There was no overall significant relationship between cumulative adherence and final symptoms.

These data indicate that accurately determining past adherence will help identify patients to target to improve their future adherence. These analyses are important for understanding time-varying measures in the clinical setting.