We found that, for the UC and the two SI groups, both the mean slope and the variance of the slope of the annual change in the pre-bronchodilator FEV1 were fairly similar to the slope and variance of the slope determined using the post-bronchodilator FEV1 (Table 2), suggesting that the post-bronchodilator measurement offered little advantage over the pre-bronchodilator FEV1 for tracking the course of COPD in LHS participants. These findings are consistent with our observation of comparable between-month variability of the pre- and the post-bronchodilator FEV1 in the SIP group ( Additional file 1).
Our results indicating that the variability of the slope of annual change in FEV1 is not substantially reduced by determining the slope based on the post-bronchodilator compared to the pre-bronchodilator FEV1 are further supported by the observation that, within each group, the proportion of participants with a statistically significant individual slope of decline in the post-bronchodilator FEV1 was similar to the proportion with a statistically significant individual slope determined from the pre-bronchodilator FEV1 since the statistical significance of the individual slope of FEV1 decline is influenced, in large part, by the variance of the slope.
On the other hand, any possible advantage, with respect to savings in time and effort, of restricting the measurement of FEV1 to only the pre-bronchodilator value for studies of the impact of an intervention on the annual rate of change in FEV1 must be balanced by the comparative size of the sample required, with adequate power, to show a significant difference between the interventions. The sample size is driven by both the effect size and the variance of the annual slope. Therefore, we determined the sample sizes needed to show a significant difference between the UC group and each of the SI groups, as well as to show significant specified differences (10, 15 and 20 ml/yr) between each of the SI groups and a hypothetical comparison group. Inconsistent differences in the required sample size were shown for determining significant differences between the UC group and each of the SI groups using pre- vs. post-bronchodilator data (Table 4). However, in general, a modestly larger sample size was required to demonstrate significance for specified differences in slope between a hypothetical comparison group and each SI group, particularly for relatively small assumed differences in slope (Table 5), thus potentially incurring an additional cost for recruitment of a somewhat larger sample size if only the pre-bronchodilator FEV1 were measured.
In studies conducted over approximately the last 25 years comparing the impact of different therapeutic interventions in COPD on the progression of the disease, it became common practice to use the post-bronchodilator, rather than the pre-bronchodilator, FEV1 for calculating the annual rate of change as the primary measure of the course of the disease [7, 9–15]. The rationale for this practice, is likely to have been based on the belief that the post-bronchodilator value better “standardizes” the FEV1 than the pre-bronchodilator value, since the pre-bronchodilator value could be affected by day-to-day and within-day variability in bronchomotor tone, as well as by residual bronchodilation from the last dose of either rescue or maintenance bronchodilator medication if an adequate washout period was not observed. The goal of this “standardization” would be to reduce the variability of the FEV1 and thus better estimate the slope of the annual change in FEV1, thereby decreasing the sample size required to demonstrate a significant difference in slope between therapeutic interventions.
On the other hand, serial spirometry studies evaluating the time course of FEV1 over 24 hrs have failed to show any difference in the circadian pattern of FEV1 comparing responses to placebo with those to bronchodilator medication [23, 24]. Moreover, the short-term response to a bronchodilator is influenced by several factors  in consequence of which the post-bronchodilator increment in FEV1 is itself highly variable both within and between patients with COPD [26, 27]. One of the factors affecting the acute response to a bronchodilator measured in terms of the absolute improvement in FEV1 is the pre-bronchodilator FEV1 % predicted, such that, across the spectrum of GOLD stages of severity from moderate to very severe, the magnitude of the FEV1 response has been observed to be largest in patients with moderate COPD and smallest in patients with very severe COPD [28, 29]. Consequently, as COPD progresses from moderate to very severe airflow obstruction over time, one would expect a progressively smaller absolute increment in FEV1 after bronchodilator administration, which could influence the annual slope of decline in the post-bronchodilator FEV1. In contrast, patients with relatively mild airflow obstruction, as observed on the baseline visit of the LHS, exhibit a minimal response to a bronchodilator [7, 30], in contrast to the much greater response in patients with moderate to severe airflow obstruction, possibly due to the effect of Poisseuille’s Law . Consequently, when such patients progress to a greater degree of airflow obstruction, one would expect a relatively larger acute response to a bronchodilator, as was demonstrated in the continuing and intermittent smokers over the 11 years of follow-up in the LHS . Whether because of these or other factors, the yearly slope of FEV1 and the variance of this slope do, in fact, differ between the pre- vs. the post-bronchodilator FEV1 has heretofore not been specifically addressed.
This study has several strengths as well as weaknesses. The major strength is the exceptional rigor with which the centralized spirometry assessments were performed and continually monitored for quality control , thus minimizing variability due to technical factors. Other strengths include the large number of subjects studied (nearly 6,000, over three-quarters of whom completed all annual visits) and the relatively high representation of females (35%) compared to most other interventional studies in COPD. A weakness is the somewhat limited spectrum of COPD represented by the subjects, all of whom had only mild to moderate airflow obstruction at entry into the study and generally had not been prescribed maintenance bronchodilator or other medication for their COPD, so that our findings might not apply to patients with severe or very severe COPD nor to nonsmokers and those without COPD. Similarly, the average age of the participants (~48 yrs) was much lower than that of COPD patients participating in pharmacotherapeutic trials. The imbalance in some of the baseline characteristics between the ~76% of participants included in the analysis and the remainder who were excluded might be another limitation. To address this limitation, we re-analyzed the data to determine the estimated annual change in FEV1 in the total LHS population of 5,887 participants using multiple imputation of the missing data (see Additional file3). The results of this analysis yielded differences between the mean slopes and slope variances determined from the pre- versus the post-bronchodilator FEV1 that were very similar to those found when the analysis was restricted only to those who completed all annual visits. Another limitation is that COPD was defined by a pre-bronchodilator ratio of FEV1 to FVC of <70%, rather than the currently recommended post-bronchodilator ratio . Consequently, some subjects with fully reversible airflow obstruction were included in the study. There were 503 subjects whose post-bronchodilator FEV1 % predicted was 90% or greater, and 1246 whose post-bronchodilator FEV1/FVC % was 70% or greater. On the other hand, subjects who were receiving regularly prescribed medication for asthma were excluded.
We conclude that serial measurements of the pre-bronchodilator FEV1 appear to be adequate for comparing the impact of different interventions on the annual rate of change in FEV1, thus simplifying the design of such longitudinal studies. On the other hand, relying only on the pre-bronchodilator measurement might require a slightly larger sample size to show significant differences between interventions, particularly if relatively small differences in slope are observed. Moreover, measurement of the response to a bronchodilator is important at baseline to exclude the presence of fully reversible airflow obstruction and, in addition, to describe the degree of partial reversibility for descriptive and potential analytic purposes, although the pre-bronchodilator FEV1 has been found to be just as accurate as the post-bronchodilator measurement in predicting mortality in the LHS . In addition, if only the pre-bronchodilator measurement is performed over time, care should be taken to ensure that subjects withhold their concomitant bronchodilator medication for a suitable washout period prior to spirometry testing. Furthermore, whether or not post-bronchodilator measurements are also performed, subjects should be studied at approximately the same time of day to minimize variability due to the influence of circadian rhythm.