Putting Health Status Guided COPD Management to the Test
The objective of the MARCH study is to study whether a treatment algorithm that is based on health status as measured by CCQ improves health status as measured by SGRQ after two years of use compared to care based on FEV1 levels as per regular (GOLD) guidelines.
This study is based on the assumption that treatment based on problems that matter to patients (as reflected in a health status measurement) will have more positive effect on their life than treatment that is based on a single measurement that has little relation with their problems (FEV1).
The selection of an appropriate primary outcome measure for the current study was an important issue during the design process. The traditional primary outcome measure in COPD research is lung function, usually represented by the FEV1. The US Food and Drug Administration (FDA) and the European Medicines Association (EMA) still routinely require this in pharmaceutical trials. However, FEV1 has been found to have a very poor correlation with markers of COPD that seem to matter most to patients, such as exercise tolerance, symptoms and also health status. Therefore, currently most researchers regard changes in patient centered outcomes such as health status, symptoms, exacerbations and functional status more important than changes in lung function. Patient centered outcomes better reflect the complexity and the impact of the disease, and several aspects of health status predict clinically meaningful outcomes in COPD. For instance, functional status as measured in health status questionnaires has been shown to predict exacerbations, hospital admissions and mortality. In most large scale COPD studies, health status is measured and demonstrated to improve after successful interventions, but it is seldom used as primary outcome. The situation is different in pulmonary rehabilitation studies where health status has been used as one of the primary endpoints.
Using health status as primary outcome measure in a study where the treatment in one arm is organized according to health status carries the risk of direct influence on the outcome. In order to reduce this potential methodological problem, a different health status questionnaire (SGRQ) is used in our study instead of the questionnaire that is used to guide the treatment (CCQ).
In the current study we decided to randomize on the patient level and not on the GP cluster level. This decision was made after careful evaluation of advantages and disadvantages of randomization on the individual and the cluster level. In this evaluation the following factors played a pivotal role. A large disadvantage of cluster randomization is the risk of selective inclusion, i.e. the physician is more likely to discover to which treatment group all his or her patients are allocated and this might, unconsciously, play a role in selecting patients for participation in the study. A second large disadvantage is the need for a much larger study population to maintain sufficient power. An additional power calculation assuming 10 COPD patients per practice, and a correlation of SGRQ within primary care practices scores of 0.14 (based on previous unpublished studies in our group), the total number of patients needed to achieve a power of 0.8 is 462. This constitutes an increase in patient number of 40%.
A disadvantage of randomizing at the individual level is the risk of contamination, loss of allocation concealment. This risk is present on both the patient level and on the physician level. On the patient level this is caused by the fact that several patients from one GP practice participate in this study and often patients in one practice know each other. Therefore patients in the control group might know patients that have been randomized into the intervention group and via that route receive information from the intervention group which they then might decide to use for themselves. However, we do not consider this to be a large risk in our study because the experimental treatment does not differ markedly from the usual care treatment, the same treatment elements are used albeit differently organized. In other words none of the patients will receive completely new and unexpected advices and therefore we expect them to conform to the recommendations given by their physicians.
The second level on which contamination might pose a risk for the study is the physician level, physicians might learn from the intervention and adjust their way of working. We try to circumvent this risk by supplying the physician with clear and individually tailored written practical advices. Physician and patients are routinely asked to report which treatment was given to each of the participants in the study giving us an accurate picture of whether or not contamination was present and if so the size of the problem.
Health care providers are not used to interpreting health status data. They need education and support to learn how to interpret the scores of health status instruments if they are to be successfully integrated into routine practice. Greenhalgh's review of health status studies concluded that information should be fed back throughout the decision making process to all clinicians involved in the patient's care and in a format they can make sense of and integrate in clinical decision making. Health status scores should therefore be presented in a coherent clinically relevant format, with clear guidelines for interpretation and preferably with to-the-point recommendations. Based on Greenhalgh's suggestions we incorporated in our study a clear treatment advice for the participating clinicians in order to avoid difficulties around the interpretations of health status scores.
Much effort was put in designing the treatment algorithms, because this is a pivotal part of the study design. During the design process choices without supporting evidence had to be made, this is because treatment based on health status is a novel concept and all previous studies were based on impairment of lung function as treatment criterion. By discussing the algorithm in different settings and with partners from various backgrounds we tried to reduce possible bias.
Vital for successful completion of the study is compliance of the care provider with the treatment advices. In the current Dutch GP practice the care for patients with chronic diseases is often transferred from the GP to the practice nurse. This applies also to implementing treatment advices. Practice nurses can achieve similar outcomes as doctors in chronic disease management. Additionally, it has been demonstrated that practices in which the organization is optimal, guidelines are better adhered to. Although this adds an extra layer in the process from measurement (lung function or health status) to effectuating the treatment, we are confident that in well organized practices with practice nurses, our advices will lead to similar results as with practices that do not work with practice nurses.
Discussion
The objective of the MARCH study is to study whether a treatment algorithm that is based on health status as measured by CCQ improves health status as measured by SGRQ after two years of use compared to care based on FEV1 levels as per regular (GOLD) guidelines.
This study is based on the assumption that treatment based on problems that matter to patients (as reflected in a health status measurement) will have more positive effect on their life than treatment that is based on a single measurement that has little relation with their problems (FEV1).
The selection of an appropriate primary outcome measure for the current study was an important issue during the design process. The traditional primary outcome measure in COPD research is lung function, usually represented by the FEV1. The US Food and Drug Administration (FDA) and the European Medicines Association (EMA) still routinely require this in pharmaceutical trials. However, FEV1 has been found to have a very poor correlation with markers of COPD that seem to matter most to patients, such as exercise tolerance, symptoms and also health status. Therefore, currently most researchers regard changes in patient centered outcomes such as health status, symptoms, exacerbations and functional status more important than changes in lung function. Patient centered outcomes better reflect the complexity and the impact of the disease, and several aspects of health status predict clinically meaningful outcomes in COPD. For instance, functional status as measured in health status questionnaires has been shown to predict exacerbations, hospital admissions and mortality. In most large scale COPD studies, health status is measured and demonstrated to improve after successful interventions, but it is seldom used as primary outcome. The situation is different in pulmonary rehabilitation studies where health status has been used as one of the primary endpoints.
Using health status as primary outcome measure in a study where the treatment in one arm is organized according to health status carries the risk of direct influence on the outcome. In order to reduce this potential methodological problem, a different health status questionnaire (SGRQ) is used in our study instead of the questionnaire that is used to guide the treatment (CCQ).
In the current study we decided to randomize on the patient level and not on the GP cluster level. This decision was made after careful evaluation of advantages and disadvantages of randomization on the individual and the cluster level. In this evaluation the following factors played a pivotal role. A large disadvantage of cluster randomization is the risk of selective inclusion, i.e. the physician is more likely to discover to which treatment group all his or her patients are allocated and this might, unconsciously, play a role in selecting patients for participation in the study. A second large disadvantage is the need for a much larger study population to maintain sufficient power. An additional power calculation assuming 10 COPD patients per practice, and a correlation of SGRQ within primary care practices scores of 0.14 (based on previous unpublished studies in our group), the total number of patients needed to achieve a power of 0.8 is 462. This constitutes an increase in patient number of 40%.
A disadvantage of randomizing at the individual level is the risk of contamination, loss of allocation concealment. This risk is present on both the patient level and on the physician level. On the patient level this is caused by the fact that several patients from one GP practice participate in this study and often patients in one practice know each other. Therefore patients in the control group might know patients that have been randomized into the intervention group and via that route receive information from the intervention group which they then might decide to use for themselves. However, we do not consider this to be a large risk in our study because the experimental treatment does not differ markedly from the usual care treatment, the same treatment elements are used albeit differently organized. In other words none of the patients will receive completely new and unexpected advices and therefore we expect them to conform to the recommendations given by their physicians.
The second level on which contamination might pose a risk for the study is the physician level, physicians might learn from the intervention and adjust their way of working. We try to circumvent this risk by supplying the physician with clear and individually tailored written practical advices. Physician and patients are routinely asked to report which treatment was given to each of the participants in the study giving us an accurate picture of whether or not contamination was present and if so the size of the problem.
Health care providers are not used to interpreting health status data. They need education and support to learn how to interpret the scores of health status instruments if they are to be successfully integrated into routine practice. Greenhalgh's review of health status studies concluded that information should be fed back throughout the decision making process to all clinicians involved in the patient's care and in a format they can make sense of and integrate in clinical decision making. Health status scores should therefore be presented in a coherent clinically relevant format, with clear guidelines for interpretation and preferably with to-the-point recommendations. Based on Greenhalgh's suggestions we incorporated in our study a clear treatment advice for the participating clinicians in order to avoid difficulties around the interpretations of health status scores.
Much effort was put in designing the treatment algorithms, because this is a pivotal part of the study design. During the design process choices without supporting evidence had to be made, this is because treatment based on health status is a novel concept and all previous studies were based on impairment of lung function as treatment criterion. By discussing the algorithm in different settings and with partners from various backgrounds we tried to reduce possible bias.
Vital for successful completion of the study is compliance of the care provider with the treatment advices. In the current Dutch GP practice the care for patients with chronic diseases is often transferred from the GP to the practice nurse. This applies also to implementing treatment advices. Practice nurses can achieve similar outcomes as doctors in chronic disease management. Additionally, it has been demonstrated that practices in which the organization is optimal, guidelines are better adhered to. Although this adds an extra layer in the process from measurement (lung function or health status) to effectuating the treatment, we are confident that in well organized practices with practice nurses, our advices will lead to similar results as with practices that do not work with practice nurses.
Source...