Article Text
Statistics from Altmetric.com
Background
The utility of patients’ self-reported asthma symptom scores in clinical trials has not been critically assessed. We investigated the types of symptom scores used in asthma clinical trials and how they are analysed and interpreted.
Methods
Systematic reviews conducted to inform asthma management guidance in England and Wales were used as an evidence base. These reviews identified 87 randomised controlled trials of the effectiveness and safety of inhaled corticosteroids and long-acting β-agonists, published during 1985 to 2006. From these trials we extracted and appraised information on the characteristics of the symptom scales employed, and the way the symptom scores were interpreted and analysed.
Results
Most (78) of the asthma randomised controlled trials (90%) reported the use of symptom scores, in all cases as secondary outcomes alongside measures of pulmonary function. Ten different numeric scales and 21 different classes of symptom combinations were identifiable among the 78 trials, which resulted in 44 unique symptom scoring scales. These scales could be further subdivided according to differences in the timing of symptom assessments and in the ways that the scores were coded. Only four (5%) of the trials used validated scales. Lung function and asthma severity typically worsen at night but in all of 11 trials that reported both day and night symptom scores, the scores were consistently lower at night, irrespective of the numerical scale, study design, and asthma interventions involved. Asthma symptom scores appeared to be ordinal measures and in most trials were analysed parametrically, without reference to any assumptions about the type of distribution or equality of the scale intervals. In the asthma trials, interpretation always focused on changes in scores and the statistical significance, rather than on actual symptoms and the significance to patients.
Conclusions
Symptom scores are widely used in asthma RCTs in conjunction with estimates of pulmonary function but they lack validation and are interpreted inconsistently and uncritically. The numerical interpretation of scores seems detached from considering the real importance of symptoms to patients. Due to the large number of unique scales in use, it is questionable whether symptom scores can be meaningfully compared across studies, as is routinely attempted in meta-analyses. Adoption of a smaller set of validated symptom scales in asthma clinical trials could assist meta-analyses, improve understanding of how numeric scores reflect patient experiences, and enable a more thorough evaluation of the quantitative properties of the scales that are used.