The percentages of participants reporting bruises a 25% of the time are shown in Figure 2. Again, clear and significant (p < 0.05) differences are found between the good compliers in the ICS group and all of the other subgroups, with > 10% of the former and only approximately 5% of the latter reporting bruises a 25% of the time. Findings were roughly comparable across the visits from 24 to 42 months.
Table 3 shows the parameter estimates and corresponding p values for predictor variables from multiple logistic regression of any bruising reported during the study for all participants, as well as for participants stratified by good and poor compliance. The results indicate a strong treatment effect, mainly among the good compliers. Older age and female gender are also significant predictors in all participants and among good compliers, but not among poor compliers. Similar results were obtained when the dependent variable was any new or worsening bruising.
Table 4 shows the percentage of participants reporting bruising at any visit by age, gender, and drug group for all participants, and for more compliant participants (ie, six or more puffs per day) and less compliant participants (ie, fewer than six puffs per day). The median age of the study population (56 years) was used as the dividing point between older and younger subjects in post hoc stratified analyses. Among all participants and those with either good or poor compliance, a significantly higher proportion of older men (ie, a 56 years of age) in the ICS group reported bruising, compared to those assigned to placebo. No differences were noted between younger men or older women in the ICS group vs the PLA group. Young compliant women showed a trend toward a treatment effect on bruising (p < 0.08 [x2 test]). Comparing the data in Table 4, we note that compliant subjects, particularly older women, reported much higher rates of bruising than noncompliant women, even in the placebo group. We are unable to explain this by-compliance difference in reported bruising among placebo-treated subjects and cannot exclude the possibility that this difference could reflect a differential recall bias in answering the questionnaire that, in turn, might have impacted the incidence of reported bruising.
Figure 2. Percentage of participants reporting bruises > 25% of the time by drug treatment assignment and compliance. Values above the bars represent the number of participants reporting bruising > 25% of the time. Differences between ICS, 6+ puffs per day, and all other subgroups were statistically significant (p < 0.05). See the legend of Figure 1 for abbreviation not used in the text.