Primary and secondary endpoints

The primary endpoint in a trial is the outcome chosen as the key measure of the effect (if any) of an intervention in a trial. It should be important to the patient, clinically relevant and directly related to the main goal of the trial. Generally, a trial has only one primary endpoint, which should be specified in the study protocol before the start of the trial. The predicted potential change in this endpoint is the basis for statistical calculations, including the sample size and the statistical power of the trial. Occasionally when it is difficult to choose a primary endpoint from several possible measures, statistical adjustments can be made when planning the trial, to allow for several primary endpoints.1 Alternatively, it may be appropriate to combine individual measures in a composite endpoint to give a more general measure of the treatment effect.1

The effects of the intervention on outcomes other than the primary endpoint (secondary endpoints), are also usually measured in trials. These should also be specified before the start of the trial. Secondary endpoints do not have the same statistical authority as the primary endpoint, and it is more likely that positive changes in secondary endpoints will be due to chance.2 Secondary endpoint results should therefore only be used to help interpret the primary result of the trial or to provide information, or prompts, for future research.

Individual endpoints

Symptoms

Asthma symptoms (e.g. cough, sputum, wheeze, chest tightness and breathlessness) measured during the day and at night (on awakening), can be assessed in terms of severity and frequency. Although symptoms have long been measured in clinical trials of asthma treatments, there is no universally accepted standard scoring system for them, and different studies record different sets of symptoms and use different scales.3 In trials, patients or carers may be asked to record symptoms daily, or weekly, on a diary card at home. Another way of recording symptoms is to use a questionnaire. For example, the Newcastle Asthma Symptoms Questionnaire, is a standardised measure of asthma symptoms used in clinical trials.4

Patients' perceptions of their asthma symptoms do not always correlate with objective measures of airways obstruction.5 Therefore, it is usual for trials to also include an objective measure of asthma control such as lung function.

Exacerbation rates

Asthma exacerbations are of great importance to patients, healthcare providers and purchasers. Exacerbations are associated with major morbidity, lifestyle disruption, hospital admission, increased cost of care and risk of death. Accordingly, exacerbations are rightly recognised as an important endpoint in clinical trials. However, there is no agreed definition of a mild, moderate or severe exacerbation, and there is a wide variation in the definitions used in trials. For example, an exacerbation has been described as a worsening of asthma symptoms severe enough to require treatment with oral or intravenous corticosteroids or a doubling of the patient's baseline intake of inhaled corticosteroid.6 Another definition is an asthma event requiring oral corticosteroids, hospitalisations or emergency visits.7 In another example, severe exacerbations were defined as those requiring treatment with oral corticosteroids or a decrease in morning peak expiratory flow (PEF) to more than 30% below baseline value on two consecutive days.8 Mild exacerbations were defined as morning PEF of more than 20% below baseline value, use of more than three additional inhalations of bronchodilator in 24 hours or awakening at night due to asthma.8 This variety of definitions can make it difficult to compare information on exacerbation rates from different trials.

Use of short-acting beta2 agonists

Recorded or recalled use of inhaled short-acting beta₂ agonist (the number of puffs during the day and at night) is often used as a marker of asthma control. However, some patients habitually take two puffs of bronchodilator, for example, before using an inhaled corticosteroid, and some use a bronchodilator before exercise to prevent symptoms. Such use must be distinguished from use of a bronchodilator for acute symptom relief ('rescue' use).

Change in corticosteroid dose

Many studies assess whether an adjunctive treatment allows the dose of ongoing corticosteroid therapy to be reduced, as a measure of treatment efficacy. For example, this endpoint has been used in trials in patients with severe asthma to assess the efficacy of inhaled corticosteroids in reducing the dose of, or avoiding the need for, oral prednisolone.9 It is also used in trials to assess whether an adjunctive treatment (such as a leukotriene receptor antagonist) allows a lower dose of inhaled corticosteroid to be used to maintain good control of asthma.10 Interpreting the results of such trials can be difficult; for instance, if the trial did not first involve stepping patients down to the minimum effective corticosteroid dose before assessing the effects of adjunctive treatment.11

Lung function

A variety of methods is used to measure lung function. The most common are forced expiratory volume in 1 second (FEV₁, the volume of air breathed out in the first second of forced blowing into a spirometer) and PEF (the maximum rate at which air is expired from the lungs when blowing into a peak flow meter or spirometer). Both these tests reflect airflow and depend on effort. FEV₁ can be difficult for patients, particularly children, to perform, and it usually needs to be measured in a clinic. PEF is easier to perform and measurements can be made at home, but they can be unreliable.12 Knowledge of PEF measurements can influence patients' self-assessment of the severity of asthma symptoms, a potential source of bias in clinical trials.13 This can be avoided by blinding the patients to the PEF reading through the use of PEF meters with a coded electronic display. Other measures of lung function are less commonly used in trials. They include forced expiratory flow in the middle part of forced expiration (FEF_25-75), an indication of airflow in the small airways.14 Different studies report lung function parameters in different ways, making comparisons between studies difficult.14

Changes in lung function measurements do not necessarily correlate with changes in other asthma endpoints. Lung function measures may therefore be an insufficient measure of treatment efficacy if used on their own. For example, in a randomised placebo-controlled trial in patients with non-steroid-dependent mild asthma, treatment with inhaled budesonide for 1 year reduced asthma symptoms but did not change FEV₁ values.15 Effects on particular asthma endpoints can differ from one drug class to another. This has implications when assessing whether the clinical trial endpoints chosen, and the efficacy differences highlighted, offer a true reflection of clinical benefit. For instance, in a randomised crossover trial involving 12 patients with asthma, those taking beclometasone reported less severe asthma symptoms than those taking theophylline, even when the PEF on each treatment was the same. Another example is a double-blind randomised placebo-controlled trial, involving 852 patients, that assessed the effects of adding formoterol (a longer-acting beta₂ agonist) to a low or a high dose of inhaled corticosteroid (budesonide 200?g or 800?g daily).8 The higher dose of corticosteroid, either alone or in combination with formoterol, was most effective in preventing severe asthma exacerbations (a primary endpoint). By contrast, in terms of effect on FEV₁ (a secondary endpoint), the combination of formoterol with either dose of corticosteroid appeared the most effective.

Lung function is frequently normal in children, and in adults with mild asthma and so lung function measures may be insensitive to change in response to an asthma treatment. In any case, clinical trials of asthma treatments often require that at entry to the study patients have a baseline FEV₁ of 50-85% of the value predicted according to age plus reversibility of airways obstruction demonstrated by an increase in FEV₁ by more than 12% after inhalation of a beta₂ agonist.17 This has the effect of excluding almost all children and adults with mild asthma and those already using inhaled corticosteroids and leads to enrolment of a very specific subset of all those with asthma. One study found that only about 15% of patients with asthma meet these inclusion criteria for a clinical trial.17

Quality of life

It is common for trials to include an assessment of patients' quality of life to complement the data from traditional endpoints such as symptom scores and lung function. Quality-of-life questionnaires aim to assess the impact of asthma on several aspects of a patient's life (including symptoms, daily activity and emotion). This seems valuable because changes in quality-of-life measures may not be reflected by changes in other asthma endpoints.18

Questionnaires must be valid (they must actually measure what they purport to assess); reliable (give the same result repeatedly for the same state in the same person); and responsive (capable of signalling a change in response to an established intervention). Two main types of quality-of-life questionnaire are used in asthma trials: generic measures, which usually give a summary of overall health benefit from an intervention and can be used across different patient populations; and asthma-specific measures, which include aspects of health that are relevant to patients with asthma. Several asthma-specific instruments are available.4 The Asthma Quality-of-life Questionnaire19 (which is also available in a version for children) and the St. George's Respiratory Questionnaire (a respiratory disease-specific measure)20 have proven validity and reliability and are commonly used in asthma trials.

Composite endpoints in asthma trials

There are several problems with using single endpoints to monitor asthma therapy. Choosing a primary endpoint from the many potential outcome measures is difficult, and the one chosen may not reflect what matters most to the patient. It can also be difficult to estimate an overall benefit from an intervention when different endpoints change to a different degree and some not at all. Also, changes in different endpoints can give contradictory messages (e.g. when symptoms improve but lung function does not change). In an attempt to overcome these difficulties, composite measures that incorporate several parameters and that aim to assess overall asthma control have been developed and are increasingly used as the primary endpoint in clinical trials.

There are many different examples of composite endpoints in published trials and no standard definition. For instance, the Gaining Optimum Asthma controL [GOAL] study was designed to assess whether patients could achieve stringent and sustained asthma control, that is, "totally controlled" or "well-controlled" asthma.7 The definitions of control were composite measures derived from the goals of treatment contained in an international asthma guideline.21 These included PEF of at least 80% of the predicted value, no rescue medication use, no symptoms, no night-time awakenings, no exacerbations, no emergency visits and no treatment-related adverse effects.7 "Totally controlled" asthma was considered to be achieved if all the individual goals were achieved for 7 out of 8 weeks. Another example of a composite endpoint is 'episode-free days', defined as days with optimally controlled asthma (i.e. no need for rescue therapy with inhaled bronchodilator therapy, an asthma-symptom score of 0, a morning PEF of at least 80% of the baseline value and no adverse effects).8 Another example is 'asthma-free days', defined as those in which the patient did not experience asthma symptoms or require unscheduled medical visits for asthma, and did not use asthma medications including pre-exercise bronchodilator therapy.22 A further example is 'asthma rescue-free days', defined as any day without asthma rescue medication and with no asthma-related health resource use.23

Markers of airways inflammation

National and international guidelines on managing patients with asthma recommend adjusting the dose of anti-inflammatory (corticosteroid) therapy according to symptoms and lung function tests.21^,24 However, the dose vs. symptoms/lung function response curve for corticosteroids plateaus at a relatively low dose of corticosteroid, indicating that above a certain dose, for most patients, there is little or no further improvement in terms of these endpoints with increasing doses of inhaled corticosteroid.25 Also, asthma symptoms do not necessarily correlate with the degree of airways obstruction15 and might not accurately reflect the underlying inflammation of the airways. In theory, if underlying inflammation is left unchecked, it could lead to remodelling of the airways in the long term.26 Consequently, there is interest in studying the effects of asthma treatments on markers that might more accurately reflect underlying inflammation. These include sputum eosinophils, exhaled nitric oxide and hyper-responsiveness to bronchoconstrictor stimuli. Studies have examined whether asthma control can be improved by using inhaled corticosteroid at doses needed to reduce sputum eosinophils27 or exhaled nitric oxide28 or to suppress bronchial hyper-responsiveness to various stimuli.29 However, the clinical relevance of these surrogate markers of inflammation is not known.

How are endpoints chosen for trials?

Endpoints should be chosen to suit the objective of the trial. An example of this was a trial designed to discover if the natural history of asthma could be changed by early corticosteroid treatment.22 The trial involved children aged 2-3 years at high risk of developing asthma who were randomised to receive inhaled fluticasone or placebo. According to the trial report's authors, the primary endpoint, 'asthma-free days' (a composite endpoint based on symptoms) was chosen because the activity of the disease was considered the most relevant to the primary objective of the study. Lung function was examined as a secondary outcome, as this was expected to contribute information about whether a decline in lung function was being prevented, an indication that early development of chronic airway inflammation and subsequent airway remodelling might have been prevented.22

For the purpose of obtaining a product licence for a drug, the choice of endpoints will be guided by regulatory authorities. For example, the European Medicines Agency recommends that the primary outcome measure for studies of bronchodilator drugs should be airway obstruction (FEV₁ and PEF).30 For anti-inflammatory therapy, it recommends placing equal emphasis on lung function and symptom-based endpoints (such as frequency of exacerbations, assessment of asthma control), with more emphasis on symptom-based endpoints for moderate and severe persistent asthma.

The choice of endpoints may have little to do with determining the place of a drug in clinical practice, not least because it can be influenced by the promotional objectives of pharmaceutical companies.

The European Respiratory and the American Thoracic Societies have recognised the need to review the endpoints used in asthma trials.31 In view of this, the organisations have formed a joint workforce to develop recommendations on standardising definitions of asthma control and exacerbations. The group is due to report its findings during 2006.

Conclusion

There are many endpoints from which to choose to evaluate the efficacy of asthma treatments. Published trials have used a wide range of such endpoints. For some (e.g. symptoms, exacerbations), there is no agreed definition. Also, different endpoints do not necessarily correlate with each other. Interpreting and comparing data from different trials can therefore be difficult. Lung function measures on their own can be an inadequate or misleading measure of treatment effect and so need to be considered with other endpoints such as symptoms, exacerbations or quality of life. Composite endpoints are being used in an attempt to estimate an overall benefit of treatment. However, at present there is no standard agreed definition of these measures. It is important to beware of over-interpretation of secondary endpoint results. Surrogate markers of the underlying inflammation of the airways are being investigated in trials, but their clinical significance is not known.

There is a pressing need for standardisation of endpoint definitions used in asthma trials. Thresholds of clinical importance for changes in endpoints should also be established.