Minimal Clinically Important Difference—Exacerbations of COPD

Abstract

Exacerbations of COPD are now recognised as being important events in the natural history of the condition and become more frequent as the disease worsens. Defining a minimum clinically important difference in exacerbation rate is fraught with difficulty. There is substantial between and within subject differences in the occurrence of these events that makes an individual evaluation of their importance problematic. At present, the most widely used definition of an exacerbation identifies an episode where the patient seeks medical help rather than a predefined change in one or more symptoms. Despite these problems, intervention studies with bronchodilator drugs, inhaled corticosteroids, and pulmonary rehabilitation appear to reduce the frequency of exacerbation events. In patients with an FEV₁ below 50% predicted there is reasonable consistency about the magnitude of change and a 4-unit improvement in the St George's Respiratory Questionnaire is commonly associated with a 20–25% reduction in the reported number of exacerbations. Individual studies vary depending upon the recruitment protocol. Patients who experience symptomatic benefit may be those in whom a clinically important change in exacerbations occurs but this concept requires testing prospectively. Existing methodologies for estimating clinically important differences are hard to apply with a binary outcome like this, and more work will be needed to develop a robust approach for dealing with this important clinical variable.

• Chronic obstructive pulmonary disease
• Exacerbation
• Health status
• Clinically important difference

Introduction

Exacerbations of COPD are an important feature in the natural history of the disease, important to both patients and their physicians and a highly relevant outcome in many intervention studies. The need to establish a minimal clinically important difference in these events is clear but this has proven to be peculiarly difficult for this endpoint. This reflects the measurement properties inherent in recording exacerbations and the element of judgmental bias between patients and payers as to what a relevant difference might be. As a result almost no work assessing the concept of MCID in the context of exacerbations has been undertaken, and hence no specific attempt has been made to relate the outcome to a template of MCID methodologies. However some crude estimates based on current reports may be possible and serve as a stimulus to a more systematic study of this important endpoint. In this article the reasons for the difficulty in applying appropriate methodologies are explored and some approaches to addressing this problem are proposed.

Many patients with chronic obstructive pulmonary disease (COPD) first seek medical attention after a period when their symptoms have increased acutely. Their first contact with a hospital is often following such an episode, which appears to define an important point within the natural history of the disease. Such events are commonly called exacerbations of COPD, a term which has now replaced the formerly widespread “acute exacerbations of chronic bronchitis.” This former usage probably explains some of the confusion that still accompanies the definition as an increase in bronchitic symptoms is not necessarily accompanied by worsening pulmonary physiology in patients without airflow obstruction. Conventional wisdom indicates that in most cases the process of exacerbation is reversible and does not produce permanentdeterioration in the underlying lung function abnormality, which is used to define disease severity. However, data from two different sources in the last few years have challenged this view. Evidence from the Lung Health Study showed that patients who attended a physician with episodes of infection were more likely to show an accelerated rate of decline of lung function over time [[1]]. Similarly, patients with more severe disease monitored in the community show a difference in the rate of decline of lung function between those with frequent exacerbations and those where these do not occur or are infrequent [[2]]. This new knowledge has led to a reappraisal of the importance of exacerbations, not only as an important clinical event in the natural history of COPD but potentially as a factor contributing to deteriorating lung function in its own right. Certainly hospitalisations due to COPD are the major cost driver of this disease [3&4] and appear to be becoming more frequent as the population ages [[5]]. Disease exacerbations are associated with poorer overall perception of health with clear evidence from several studies that baseline health status is lower in groups of patients with similar pulmonary function among those who exacerbate most often [[6]], that the rate of deterioration of health status over time is greater in those were exacerbations are most frequent [[7]], and this appears to explain much of the effect of some treatments on baseline health status and its subsequent decline [7&8]. More recently, studies of health status change after exacerbations of disease have shown that, far from being brief in duration, disease exacerbation have a lasting effect on the patient's perception of ill-health [[9]]. Much of the acute impairment of health status resolves in the 4 weeks after an acute exacerbation, but there is a further clinically important change over the next 8 weeks and there are still statistical significant differences extending out to 6 months compared with the 12-week values, suggesting a long-lasting effect for these episodes. Moreover, those individuals who exacerbate again before their first episode has resolved appear to have a permanent impairment of health status [[9]]. Not surprisingly, patients are concerned about these episodes and as part of a validation process for the COPD control questionnaire, a document developed to help primary care physicians evaluate patient symptoms more rapidly it became evident in talking to patients that one of their major concerns was their inability to control breathlessness when the disease exacerbated [[10]].

Thus from the perspective of the patient, the physician, and the payer, exacerbations of COPD are an important clinical entity and their prevention and effective treatment is a major goal of all COPD management programmes [11&12]. Inevitably, the assessment of COPD exacerbations has become part of the range of outcomes considered when any new treatment is evaluated. To date, no therapeutic intervention has been completely successful in abolishing exacerbations in all COPD patients to whom it has been offered. The changes in exacerbation frequency have varied between 17–25% of that seen with placebo therapy leading to the reasonable question of how large must the reduction in the number of exacerbations be before a COPD patient thinks it is worthwhile and hence, what is the minimum clinically important difference in this outcome measure. At the same time, there is a regulatory and commercial expectation that clinical trials will quantify the number and severity of exacerbations. This approach has identified a series of problems, which make COPD exacerbation one of the hardest outcomes to express in a way that can be compared between studies.

Defining an Exacerbation

Both patients and physicians are very familiar with what they mean by an exacerbation. However, precisely defining an exacerbation has proven to be remarkably difficult. Initial definitions drew heavily on studies of patients complaining of chronic bronchitis, a condition that is not necessarily associated with airflow obstruction [[13]] but can be accompanied by worse health status [[14]]. The most influential study using a definition based on bronchitic symptoms was a multi-centre Canadian Medical Research Council trial of the effects of antibiotics in episodes believed to be infective in origin [[15]]. This study divided symptoms into major and minor ones, the major symptoms being an increase in cough, sputum production, sputum purulence and breathlessness. This approach has been adapted for the patient's participating in the large East London, UK, longitudinal cohort study key features of which are shown in Table 1. An alternative approach more readily applied in the clinical trial setting, has led to the development by a group of clinical experts of the ‘consensus’ definition of a COPD exacerbation that states that an exacerbation is “A sustained worsening of the patients condition from the stable state and beyond normal day-to-day variation that is acute in onset and necessitates a change in regular medication” [[16]]. A need for some form of health carecontact to be present is useful when patients are being seen intermittently, as in a clinical trial setting and are recalling significant clinical events. On the other hand, this definition is clearly dependent upon the culture and health-care system in which the patients live and takes no account of episodes where the patient manages symptoms independently and decides not to seek medical help. There are good data from the East London studies that patients do indeed do this, and as a result the number of exacerbations recorded in that study is almost double that seen in many other large trials of apparently similar patients [[6]]. The consensus definition does not give any indication of the duration of the event and although there is some circumstantial evidence to suggest that patients who receive oral corticosteroids are perceived as having worse episodes [[17]], this may only apply to those living in Western Europe rather than in other parts of the world.

Table 1.  Criteria Used to Identify an Exacerbation of COPD from Daily Diary Cards as Used.

•Major symptoms

-Increased cough

-Increased sputum volume

-Sputum purulence

-Dyspnea

•Minor symptoms

-Wheeze

-Fever

-Cold symptoms (nasal congestion/discharge)

•Acute exacerbation (duration of 2 days)

-2 major symptoms

-1 major symptoms + 1 minor symptom

•Change

-Better

-Worse

-Same

•Severity

Number of symptoms

From Refs. [[2]], [[5]], and [[17]].

Attempts have been made to capture exacerbations prospectively and thereby overcome the problems of patient recall. To do this, patients must fill in a daily diary card, but the questions that should be included and the scaling system to be adopted have not yet been agreed upon. How episodes identified in this way relate to those relying on the consensus definition is not known. How the symptom based event is affected by the method used to collect the data (electronic or manual diary card, symptom intensity scale or perceived change in symptom intensity) is equally unclear. To date there has been no systematic examination of the occurrence of episodes of increased bronchitic symptoms in a population defined by abnormal pulmonary function, making it hard to relate the newer definitions to those used previously to denote an acute exacerbation of bronchitis. So far the East London COPD study has led the way in reporting criteria for the presence of a new event and indicating a standard of severity for that exacerbation their criteria being listed in Table 1.

They have used a comparator system in which patients relate their current clinical state to their memory of what they were like on the preceding day. There is some precedent for using this approach but it also has disadvantages. There is an effective ceiling on symptom intensity and so in this study patients were considered to have more severe exacerbations when more symptoms are reported rather than the symptoms themselves being considered slightly, or much worse than normal using a Likert scale to grade of their severity [[18]]. Data from these studies also illustrate that patients do not, unlike their doctors perceptions, always report increases in all symptoms with every exacerbation. Over 60% of exacerbations are associated with an increase in the sensation of breathlessness. This still leaves a substantial number of episodes in these otherwise symptomatic patients were breathlessness is not a driving symptom of an apparent period of deterioration [[18]]. Moreover, recording data in this way allows statements about the prevalence of symptomatology in a population to be made rather than describing change in an individual patient. Finally, the need to commit to recording symptoms on a daily diary card is probably acceptable for a carefully monitored cohort of patients where there is regular telephone access to the coordinating centre. It is unlikely to be applicable beyond the research setting.

One further problem should be considered. Patients with COPD have been studied across a wide range of disease severity from people who are only mildly symptomatic and have mild to moderate degrees of airflow obstruction [[19]] to individuals, whose mean FEV₁ is around 50% of predicted [[20]] or even as low as 35% predicted [[21]]. In all of these studies the patients were followed for at least one year and exacerbations have been reported. It is clear that the number of exacerbations involving healthcare contacts rises as the patients become sicker and this may be influenced by treatment. This has been specifically explored in a further analysis of the data contained in the ISOLDE and here it was evident that patients with an FEV₁ below 50% predicted were much more likely to report either no exacerbations or only one per year follow-up (Figure 1) [[22]]. It should not really be surprising that the effects of the intervention, in this case an inhaled corticosteroid, were much easier to demonstrate as far as exacerbations were concerned in individuals with more impaired pulmonary function in whom exacerbations were more frequent. This may be more an issue of statistics than biology. This variation in incidence with disease severity of an intermittent state (exacerbation) makes this outcome very different from the others considered in the articles which accompany this one.

Figure 1. Distribution of exacerbations of COPD defined by the consensus definition as described in the text during the 3-year follow-up period of the ISOLDE study. Data are separated into events occurring in patients with an FEV1 greater than 50% predicted (mild) and those below this (moderate–severe). Reproduced with permission from Ref. [[22]].

It is also clear from Figure 1 that the annualised exacerbation rate itself is not normally distributed. Morerecent studies have modelled exacerbation rates using a Poisson distribution that gives a normalised mean. Unfortunately, this does not allow us to look at more conventional variance measures like standard deviation or standard error of the mean that have been used to derive an MCID in the accompanying papers. Indeed, exacerbations as an outcome are difficult to express in statistically normalised ways. From the patient's perspective, you either have an exacerbation or you don't, rather like being pregnant. This is certainly true of the way in which the consensus definition operates and is also true with the diary card approach used in the UK cohort. Until we can develop better ways of grading the severity of an exacerbation and developing an evidence-based definition of an exacerbation it will be difficult to conduct either expert based or patient based consensus evaluations of this outcome.

Estimating a Minimally Clinically Important Difference

From the foregoing it is clear that techniques developed to study symptoms and health status are not easily applied to an exacerbation outcome. Nonetheless, there is a surprising degree of agreement between different clinical investigations about the magnitude of treatment effects seen when this outcome is studied. All of the studies listed in Table 2 have used consensus definition of exacerbation or something closely allied to it, e.g., the number of hospitalisations in the population of patients undergoing pulmonary rehabilitation [[23]]. All of them were randomised and controlled trials lasting for one or more years. In all cases the measurement of health status was also made using the same instrument, the St. Georges Respiratory Questionnaire. As this is one of the health status tools in which most work has been done to establish an MCID and as noted above, differences in exacerbation contributes significantly to changes in health status, then we can gain some insight into what an important difference might be. The trials listed here are relatively substantial and the comparisons reported include different treatment limbs from the same trial, the numbers referred to representing the sum of patients in each comparison limb. In general terms active therapy reduced exacerbations by between 20–25% of the rate seen in similar patients randomised to placebo. Change in health status varied from just less than 2 to 7.5 units with a grand mean change close to 4 units. Thus, as a very broad generalisation of the data in Table 2 the minimally significant change in health status (i.e., 4 units on the SGRQ) was associated with a fall in exacerbation rate of just over 22%. This is an example of a crude anchor-based approach to the estimate of MCID (Table 3).

Table 2.  Exacerbation Rates and Accompanying Change in Total St. Georges Respiratory Questionnaire Score (SGRQ) in a Series of Randomised Clinical Trials of at Least 1 Year's Duration.

Trial	Therapy	n	% Reduction	Δ Health status SGRQ units
Casaburi [[23]]	Tio/plac	844	20	3.4
Vincken [[24]]	Tio/ipra	490	20	3.3
TRISTAN [[25]]	Salm/plac	733	20	1.9
TRISTAN [[25]]	FP/plac	735	19	2.2
TRISTAN [[25]]	Comb/plac	719	25	4.3
Szafranski [[26]]	Bud-form/plac	413	24	3.9
Calverley [[20]]	Bud-form/plac	510	24	7.5
Griffiths [[22]]	Rehab/plac	200	26	4.8

Plac = placebo; Tio = tiotropium; Salm = salmeterol; FP = fluticasone propionate; Comb = combination of salm and FP; Bud = budesonide; Form = formoterol; Rehab = rehabilitation. Figures in superscript refer to source references.

Table 3.  Possible MCID Data for Exacerbations of COPD.

Name of outcome measure: Exacerbation of COPD

Range: 0–26 events per year

SEM: not known

Reliability: unknown and likely definition-dependent

Effect Size: unknown

Distribution-based MCID: not available

Anchor-based MCID: 22% difference from baseline (consensus definition based)

Expert physician-based MCID: not yet defined

Patient-based MCID: likely influenced by intensity of baseline symptomatology. One event per year is the current best estimate.

SUGGESTED MCID:

An individual reduction of one event per year is important. In clinical trials, a 20% reduction in overall exacerbation rate compared to placebo is a reasonable effect.

This approach is, to say the least, not very robust and has several important weaknesses within it, apart from the obvious generalisation which pooling mean data in this way represents. It is clear from some of the trials that improvements in health status were not necessarily wholly due to changes in exacerbation frequency. Thus, in the studies using tiotropium, a long-acting inhaled anticholinergic bronchodilator drug, significant change occurred in the patient's perception of breathlessness and this clearly could influence their overall well-being [24&25]. This was not assessed directly in the trials of combined inhaled corticosteroid and beta-agonist therapy [[21]][26&27]. Shorter-term studies using combination drugs suggested that important improvements in breathlessness defined by a one unit or greater improvement in the TDI could accompany treatment with long-acting inhaled bronchodilators alone or in combination with an inhaled corticosteroid [24&25][[28]], the diary card scoring of breathlessness in these trials also showing statistically significant improvements. The entry criteria forinclusion in the studies differed, particularly with regard to the degree of bronchodilator reversibility and the time since the last clinical exacerbation. It is quite possible that different treatments affect different aspects of health status. These do not necessarily translate or run in parallel with how they affect exacerbations. There is a suggestion from the published data and some as yet unpublished studies, that long-acting β-agonists have more impact on health status than they do on exacerbation rates and this was certainly true in the studies of budesonide and formoterol [[21]][[27]]. Finally, the largest change in health status seen was moderately large mean improvement of 7.5 units in the SGRQ, This was in part due to the run-in period when patients were exposed to intensified treatment with oral corticosteroid and inhaled formoterol for two weeks before randomisation to placebo or active therapy [[21]]. To date, this is the only trial that has used this approach but it is possible that this variation in design contributed to the larger than anticipated change in health status.

Thus, although there are real concerns about aggregating information in this way, at present this is the only method we have to gain any impression of whether the magnitude of change in exacerbation frequencies seen with therapy translate into an important reduction in disease impact for the patient. Whether this mean change in exacerbation frequency is the average effect of current treatment or the minimally acceptable clinical difference is, of course, not yet addressed.

Future Work

For the reasons already explained, exacerbations are a particularly difficult area in which to determine what a minimally clinically important difference might be. Much more work is going to be needed to anchor changes which occur during an exacerbation to the patient's own perception of these episodes. If we could be confident that there was a good correspondence between a diary card designated exacerbation and the patient's own rating of the episode then consensus approaches between groups of patients might yield a more realistic estimate of what a minimally clinically important difference would be from the patient's point of view. Simply asking individual patients what an important difference would be to them produces the answer that they wish to have no exacerbations at all at any time and hence a more meaningful question might be, how much trouble would they like to be put to if we could abolish your exacerbations? This approach has been used elsewhere in establishing clinically important differences and may well be the most useful one in this area. A further difficulty is recall bias especially in those individuals were the events are less frequent. Thus you may not get the same answer if the patient is asked to rate the importance of the event 4 months after it occurred compared with 4 weeks post-exacerbation.

It would clearly be of great importance to ensure that when patients try and evaluate this outcome they are at least at a similar stage of COPD. Unlike exercise performance, which can be observed by patients and discussed in terms of its limitation on their lives, the relatively infrequent and apparently clinically milder exacerbations that occur in the early stage of the disease are not necessarily the same either in their consequences or the degree of health status disruption as episodes that occur in patients whose lung function has deteriorated more. Thus, an upper respiratory tract viral infection that might produce a few days of discomfort and ill health in somebody whose FEV₁ is 65% predicted could easily lead to hospitalisation and a period in an intensive care unit for somebody whose FEV₁ is 25% predicted. Comparing like with like will be essential in developing approaches in this area.

Although this is a particularly difficult topic, it should not be beyond us to address it. Perhaps new methodologies will be needed to do this effectively. Much work is needed because despite all the methodological problems, exacerbations are a key event in the COPD patient's life and one whose prevention remains a priority.