We have shown that, in moderate to very severe COPD, mean responses to a near-maximal bronchodilator challenge decline progressively over time to a statistically significant, albeit modest and highly variable, extent, irrespective of the method of calculating the responses (absolute, relative or % predicted pre-post bronchodilator change). These downward trends in bronchodilator responses were observed in nearly all subgroups defined by the initial severity of airflow obstruction, age, gender, smoking status over the course of the study and use of ICS at baseline. However, the magnitude of the decline in responses for FEV1 or FVC differed significantly within some of these subgroups; for example, declines over time in absolute, relative and % predicted FEV1 and/or FVC responses were larger in patients with severe and very severe vs. mild/moderate airflow, > vs. ≤65 years of age, in sustained ex-smokers than continuing smokers and in patients receiving versus not receiving ICS at baseline (Tables 2, 3 and 4). The declines in responses for FVC tended to be larger than those for FEV1 when these were assessed as absolute changes, but not as relative or percent predicted changes. While some differences in the mean annual changes in FEV1 and FVC responses were noted between the placebo and tiotropium arms of the trial, these differences were not statistically significant irrespective of the method of expressing the bronchodilator response.
Responses to a bronchodilator in COPD patients are well known to vary over a relatively short time frame such that a large proportion of patients who respond significantly to a bronchodilator challenge on one day fail to do so on another day and vice versa over a relatively short time frame [4,5,15]. On the other hand, long-term trends in bronchodilator responses over more than one year have infrequently been measured [7,8]. In the IPPB trial, in which the average baseline pre-bronchodilator FEV1 (36.1% predicted) was comparable to that in UPLIFT (39.4% predicted), the mean change in the relative FEV1 response per year over 3 years (-0.58/yr) was similar to that which we observed in UPLIFT over 4 years (-0.68/yr) but, unlike the present findings, was not significantly different from zero . On the other hand, the change in % predicted FEV1 response over 3 years (-0.36/yr) was both similar to that noted in the UPLIFT population over 4 years (-0.33/yr) and also significantly different from zero. The long-term change in absolute responses in the IPPB trial was not reported.
In contrast, in the 5-year LHS, in which the mean baseline pre-bronchodilator FEV1 (75.4% predicted) was much higher than that in either the IPPB trial or UPLIFT, a substantial increase in responsiveness (assessed as relative, absolute and % predicted responses) was noted over the first year, with either a slight further increase or no change over the ensuing 4 years and substantial increases from year 5 to year 11, except for the absence of any perceptible change only in absolute responses in sustained quitters over this time frame . The reason for these disparate findings in the LHS compared to both the IPPP study and UPLIFT is unclear but might possibly be related to the much milder degree of airflow obstruction in LHS participants compared to those in the other two studies. The observation that the decline in relative responses in UPLIFT was significantly greater in patients with severe/very severe than mild/moderate airflow obstruction seems consistent with this possible explanation; however, even the subgroup of UPLIFT subjects with mild/moderate obstruction showed highly significant declines in both FEV1 and FVC responses, irrespective of the method of expressing these responses.
In this analysis, as in previous reports [5,7–9], the response to a bronchodilator has been shown to be highly variable both between and within individuals. Moreover, because of this variability and because responsiveness has not been shown to be predictive of exacerbations or mortality in ECLIPSE  or predict the long-term response to a bronchodilator over 1 year [16,17], bronchodilator responsiveness has been considered to be an unreliable phenotype . However, the progressive decline in bronchodilator responses over time demonstrated in UPLIFT, as well as in the IPPB trial, mirrors to some extent the usually progressive, but admittedly variable, decline in lung function characteristic of COPD, suggesting that these two phenomena might be inter-related. One can only speculate as to the mechanism of the observed declines in bronchodilator responses over time. One possible mechanism is a progressive increase in the thickness of the walls of the small airways with progressive increases in the severity of airflow obstruction, as reported by Hogg et al. ; the resulting decreases in airway wall compliance could diminish the effect of bronchodilator-induced airway smooth muscle relaxation in increasing the patency of the lumen. It is also possible that the age-related loss of lung elastic recoil  that is most likely amplified in patients with progressive emphysema could counteract drug-induced bronchodilation and reduction in air-trapping by increasing dynamic airway compression.
One clinical implication of the progressive decline in bronchodilator responses over multiple years is that this might result in a reduced effectiveness of bronchodilator therapy on clinical outcomes in COPD as the disease progresses over time, at least in some patients in view of the large inter-individual variability observed in the decline in responsiveness. The finding of a modest but statistically significant within-individual relationship between declines in acute bronchodilator responses on the one hand and worsening SGRQ scores on the other suggests that decrements in the response to a bronchodilator over time might be associated with poorer clinical outcomes, although this association does not necessarily imply causality. Moreover, the only modest changes in SGRQ score in association with declines in bronchodilator response (1.2 unit increase in SGRQ for each 10 ml decline in acute FEV1 response) argue against a clinically meaningful relationship. Another implication of our findings is that the decline in the acute bronchodilator response over time would lead to a partial convergence of the slopes of decline in lung function calculated from the pre- and post-bronchodilator FEV1 (and FVC), resulting in a steeper post- than pre-bronchodilator slope . Moreover, these potential consequences are likely to be relatively independent of the severity of airflow obstruction, age, gender and smoking status since significantly progressive declines in bronchodilator responses were seen in most of these subgroups. The possible impact on the slope of change in post- vs. pre-bronchodilator lung function over time needs to be taken into account in the design of long-term trials in which the annual decline in post-bronchodilator lung function is measured as a means of assessing the rate of progression of COPD.
The strengths of the present analysis include the large number of patients with varying degrees of severity of airflow obstruction who were followed over an extended period of time, the high quality and reproducibility of spirometry  and the relatively large doses of the two different classes of short-acting bronchodilators that were administered along with the timing of post-bronchodilator spirometry to coincide with the time of expected peak action of each class of bronchodilator.
A major limitation is the large drop-out rate with only 27%, 35% and 40% of subjects completing visits at 2, 3 and 3¾ yrs, respectively. The current analysis uses all available data and the statistical inference from linear mixed effects models is valid under the missing at random assumption. Moreover, those who discontinued the trial prematurely were more likely to have fared poorly during the trial, suggesting that their bronchodilator responses, had they been measured subsequent to their withdrawal, might tend to be even less robust than the responses at later time points in those subjects who completed the trial.
In conclusion, acute responses of both FEV1 and FVC to near maximal doses of two different bronchodilators, while considerably variable both between and within individuals, on average diminish progressively and significantly over time, consistent with the usually progressive decline in lung function with age in patients with COPD. These declines were independent of the method of expressing the bronchodilator response and tended to be larger in patients with severe/very severe compared to those with mild/moderate airflow obstruction, in patients >65 years of age and in former than continuing smokers and in those not on ICS at baseline. These declines in the response to a bronchodilator imply a possible diminution in the clinical efficacy of bronchodilator therapy over time and may account for differences in the slopes of lung function decline with age when calculated using the post- compared to the pre-bronchodilator value.