Discussion
This study is the first to examine in a large dataset the factors associated with changes in the exacerbation frequency category of patients with COPD. We had hoped to identify a parameter that might prospectively alert physicians to an imminent change in category but no single parameter emerged. Those variables that did change in the year before IE patients became frequent exacerbators (falls in FEV1 and 6-minute walking distance) or FE patients switched to IE (fall in platelet count) were too small in magnitude relative to the overall variability to provide a useful predictive tool. We did observe more hospitalized exacerbations in those changing from IE to FE, and fewer hospitalized exacerbations in those changing from FE to IE, compared to those patients who remained in their category. Since, the number of hospitalized exacerbation over year 1 partly determined how the patients were categorized; this parameter could not be considered sufficiently independent or prospective for predicting a change in category. However, it is possible that the presence, or absence, of exacerbations severe enough to warrant hospital admission indicates a transition in the trajectory of the disease. Our results highlight the multi-facetted nature of the disease and suggests that all patients are at risk of developing the frequent exacerbator phenotype.
Across the whole study population, exacerbation frequency was unchanged in year 2 compared to year 1. This agrees with another observational study that found no change in exacerbation frequency over 6 years of follow-up albeit with a symptomatic definition of exacerbation. Aaron and colleagues have reported with transitional regression models that patients experience a significant acceleration in the rate of exacerbations over just one year. It might be that changes in inter-exacerbation intervals provide a more sensitive measure of changes in exacerbation frequency over time, but this metric lacks the simplicity required for clinical assessment. We observed that 44% of patients had no change in exacerbation frequency between years 1 and 2. In the remaining 56% of patients, the changes in frequency had a bell-shaped distribution with an approximately equal number of rises and falls. This could simply be "regression to the mean" but there will also be some random variability in the number of viruses and bacteria circulating in the community which are thought to trigger the majority of COPD exacerbations.
The rise in exacerbation frequency that we observed in the IE group could involve four factors. Firstly, our definition of IE as 0 or 1 exacerbation per a year meant that falls in exacerbation frequency were restricted to a maximum of -1 per year, but there was no limit to any rise in frequency, and the maximum increase we observed was +10 per year. Secondly, COPD progresses and higher exacerbation frequencies are more likely to be reported in more severe patients. We found that the patients who changed from IE to FE were generally more severe than those who stayed in the IE group over a wide range of patient characteristics. Thirdly, some patient characteristics fell over the year preceding the change from IE to FE. An acute deterioration is occasionally observed in clinical practice but the underlying reasons for the change are not totally recognised. The physiological markers that worsened relative to the normal reduction observed in those patients who did not change category were FEV1 and 6MWD. These are important indicators of COPD severity, with rapid decline in both FEV1 and 6MWD associated with increased mortality. Unfortunately, the magnitude of the falls in FEV1 and 6MWD were small compared to their overall variability and thus would not make useful targets for monitoring risk or early intervention. Fourthly, it is possible that exacerbation frequency changed because patients started to seek treatment for previously unreported exacerbations or increased their adherence to already prescribed treatments. About 50% of exacerbations in patients with moderate to severe COPD are unreported. Unreported exacerbations are generally similar to reported exacerbations in symptoms and spirometric changes and contribute to a poorer quality of life but there is much heterogeneity of reporting behaviour and even some patients with severe airflow obstruction will not report exacerbations.
We observed that a significant falls in platelet count occurred over the year proceeding patients changing from the FE to IE group but the standard deviation of these changes was too large to be of clinical utility. Platelet count has been found to be elevated in COPD compared to healthy controls and increases during lower respiratory tract infection. Platelets are not only involved in haemostasis but produce a broad array of inflammatory mediators and factors that alter innate and adaptive immunity. We speculate that a reduced platelet count is indicative of reduced airway inflammation or improvement in immunity to infection thus reducing susceptibility to future exacerbation. There is also emerging evidence that statins, which reduce platelet-activation, have the potential to reduce exacerbation frequency in COPD.
The relatively large proportion of patients changing from FE to IE could also be explained by the method by which exacerbation frequency was assessed. A number of patients could be experiencing three exacerbations over a two year period, with for example, two exacerbations in year 1 and then one exacerbation in year 2. They would initially be categorized as FE but change to IE in the following year. Another possibility is that that patients who normally experience two exacerbations per year change to having 3 in one year and 1 in the next year, because they have an exacerbation just before, rather than just after, the division between years. Another explanation is that participation in an observational study reduces anxiety and/or improving adherence to treatment with a consequential fall in exacerbation frequency.
There are a number of limitations concerning the major findings that should be mentioned. A practice 6MWD test was not performed and approximately 6% of the results might have differed if this rehearsal had this been undertaken. The declines observed in FEV1 in the Eclipse study were modest but in-line with a recent interventional study. Measurements of FEV1 and blood cell counts may have affected by exacerbations that took longer than 30 days to resolved, by the development of co-morbidities or initiation of pulmonary rehabilitation. The analysis was based on change in the total number of exacerbations and we did not examine the data by change in the number of hospitalized or oral corticosteroid treated exacerbations.
Amongst the strengths of this study was the follow-up of a large number of COPD patients with systematic measurements on a wide variety of patient characteristics. The patients were recruited worldwide from primary and secondary care and thus the results are generalizable to clinical practice in COPD. We did not feel that changes in oral or inhaled medication were responsible for a change in exacerbation phenotype. Our reasons were the very small proportion of patients involved and the impossibility to determining whether any change in treatment was in response to a rise in exacerbation frequency or to pre-empt an increase.