Epidemiology and costs of chronic obstructive pulmonary disease

Summary

Worldwide, COPD is in the spotlight, since the high prevalence, morbidity and mortality present challenges for healthcare systems. The inconsistent use of terminology for COPD, and the lack of widely accepted diagnostic standards for the diseases that are included in the coding within the COPD spectrum have contributed to the inaccuracy of mortality data for COPD. The approach of using the single term COPD (rather than individual coding for chronic bronchitis, emphysema and chronic airway obstruction) is being championed in global and national guidelines, with the expectation that this will improve awareness, simplify the coding and, ultimately, lead to greater accuracy in death certification for COPD. In 2000, there were more deaths in the USA from COPD among females than males. In contrast to the trend for cardiovascular diseases, death rates from COPD have been rising steadily over the past few decades. This striking increase in COPD as a cause of death is projected to occur because of the worldwide epidemic of smoking, and the changing global demographics where more people in developing countries are living longer and, therefore, are at risk of COPD for longer. Prevalence estimates may vary widely depending on which spirometric criteria are used. The BOLD initiative, which is a programme designed to provide strictly standardised, rigorous and practical methods for estimating the prevalence and social and economic burden of COPD, is described.

Introduction

Attention is being focused on COPD worldwide because it has become clear that it is an under-recognised, under-diagnosed and under-treated disease. COPD is estimated to become the third leading cause of death worldwide by 2020. The increase in mortality and morbidity from COPD is occurring worldwide as a result of both the epidemic in tobacco use and, especially in developing countries, the changing age structure of populations. This trend will continue, even if all efforts to stem the tobacco epidemic are successful. In the USA and Canada, the number of COPD deaths in females recently outnumbered the number of deaths in males. This statistic demands attention, since COPD has traditionally been thought of as predominantly affecting males. It is likely that this trend will soon be seen in other Western countries, since females are living longer and have smoked in increasing numbers since about 1940.

COPD prevalence is generally higher than is recognised by health authorities 4, 5. Few population-based prevalence surveys have been carried out, and prevalence estimates have often relied on expert opinion or self-reported doctor diagnosis, a notoriously unreliable source of information for COPD. For example, in the USA National Health and Nutrition Examination Survey III, 70% of those with airflow obstruction had never received the diagnosis of COPD 7. The IBERPOC study in Spain also reported that there was no previous diagnosis of COPD in 78% of identified cases and, even more worrisome, only 49% of those with severe COPD were receiving some kind of treatment for COPD 8. Recently, the Nippon COPD Epidemiology (NICE) study in Japan, also presented within the current series, had a similar finding 9. During the 1990s, asthma surveys successfully identified huge variations in asthma prevalence, both in children and adults, as high as 20-fold. It appears that the geographical distribution of COPD is more homogeneous than asthma, at least in the developed countries. It seems likely that the distribution of COPD follows the distribution of its risk factors very closely, of which smoking is undoubtedly the most important worldwide.

COPD is in the spotlight worldwide, as the high prevalence, morbidity and mortality present challenges for healthcare systems. From the patient's perspective, it is also a disease that has a profound effect on quality of life 10. The burden of COPD can be assessed in a number of ways, including the following: mortality; morbidity; prevalence; disability-adjusted life yrs; cost; and quality of life. A number of authors have comprehensively reviewed this topic in detail elsewhere 11, 12. Due to space restrictions, this paper focuses on mortality, morbidity and prevalence, with a particular emphasis on prevalence.

Worldwide mortality and morbidity from COPD

Of all of the descriptive epidemiological data available for COPD, mortality data are the most readily accessible. Inconsistent use of terminology for COPD, and the lack of widely accepted diagnostic standards for the diseases that are included in the coding within the COPD spectrum have contributed to the inaccuracy of mortality data for COPD 4, 5. For example, the International Classification of Diseases (ICD)-9 codes used for COPD include: chronic bronchitis (ICD-9 491); emphysema (ICD-9 492); and chronic airway obstruction (ICD-9 496). Asthma (ICD-9 493) should not be included in the COPD definition. The approach of using the single term COPD is being championed in global and national guidelines, with the expectation that this will improve awareness, simplify the coding and, ultimately, lead to greater accuracy in death certification for COPD. For the next iteration of the ICD coding, the consolidation of emphysema, chronic bronchitis and chronic airway obstruction will be an important step towards obtaining more accurate data on the distribution of COPD worldwide.

Death rates from COPD have been rising steadily over the past few decades. This trend is particularly striking, since it is opposite to the trend for cardiovascular diseases, the most common chronic diseases. In the period of 1965–1998, death rates from coronary heart disease in males in the USA dropped 59%, and deaths from strokes and other cardiovascular diseases decreased 64% and 35%, respectively. Over the same period, deaths from COPD increased by 163% 4, 5.

COPD has traditionally been thought of as a disease of elderly, smoking males. It was, therefore, surprising to see that, in 2000, there were more deaths in the USA from COPD among females than males 13. Although the rates of death were still higher in males than in females, it reflected the different age structure of the USA population for both sexes, with females living longer and, therefore, being more at risk of developing COPD. This dramatic change in the sex distribution of mortality is likely to be seen in other countries that have been lagging behind the USA in smoking patterns among females, but have been catching up over the past few decades. It is worth noting that in all countries but three (Norway, Sweden and New Zealand), and in these ones only since 2003, females have never smoked as much as males 14.

COPD mortality is not only increasing in developed countries. Worldwide, COPD was the sixth leading cause of death in 1990, and presently is the fifth. The Global Burden of Disease (GBD) Study projects that, by 2020, COPD will become the third leading cause of death worldwide 2. The methods of the GBD Study, and its COPD results, are discussed in this issue in detail 15.

This striking increase in COPD as a cause of death is projected to occur because of the worldwide epidemic of smoking and the changing global demographics, with more people in developing countries living longer and, therefore, being at risk of COPD for longer. A recent United Nations report projected that the number of people worldwide aged >60 yrs will nearly double over the next 50 yrs, and, by the mid twentieth century, the population aged >100 yrs will be 15 times higher than today 16. It is actually the changing demographics worldwide that is driving the COPD tidal wave even faster than the increase in smoking worldwide 14. Not surprisingly, there are large differences across countries in COPD mortality rates, as illustrated in the recent ERS European Lung White Book 17. Given the well-established inaccuracy when coding COPD as a cause of death, these figures must be taken as rough estimates. Males consistently have higher COPD death rates than females in all countries.

Morbidity assessment includes physician visits, emergency department visits and hospitalisations. COPD morbidity data are often less reliable than mortality, since the various ways of measuring morbidity are more prone to external factors, such as the availability of hospital beds, local and regional use of filters from primary to secondary care, the coding for utilisation being affected by reimbursement patterns, and other such potentially biasing factors. Despite these external factors, morbidity for COPD is important to track, since these data can provide an estimate of the need for health services 10, 11.

COPD prevalence

Whatever the disease, prevalence estimates depend on the definition that is used for diagnosis. For COPD, a number of different approaches have been used, including the following: doctor diagnosis; diagnosis based on the presence of respiratory symptoms; and a diagnosis based on the presence of airflow limitation (without or with a bronchodilator test as recommended by GOLD). These different approaches, not surprisingly, give very different estimates, with doctor diagnosis giving the lowest estimate of prevalence 18, diagnosis based on respiratory symptoms giving the highest estimates, and a diagnosis based on spirometry giving an intermediate estimate 13. Since the GOLD guidelines were published, the need for spirometry in making the diagnosis of COPD has been generally accepted, and this has now become the “gold standard”, at least for epidemiology. However, even when using so-called objective measurements, estimates may vary widely depending on which spirometric criteria are used 19, 20. GOLD recommends that a post-bronchodilator forced expiratory volume in one second (FEV₁)/forced vital capacity (FVC) <70% confirms the diagnosis of COPD, and FEV₁ provides a way to stage COPD. For instance, stage 1, or mild COPD, is defined as a post-bronchodilator FEV₁/FVC ratio <70% together with an FEV₁ >80% predicted, all based on predicted (or reference) values based on sex, age and height (table 1⇓). The same spirometry thresholds have recently been adopted by the 2004 ERS/ATS COPD guidelines 6. Use of other criteria, such as a pre-bronchodilator FEV₁/FVC ratio or FEV₁ % predicted without the FEV₁/FVC ratio, or other thresholds, will give very different estimates 21.

An example of the importance of using spirometry to estimate the prevalence of COPD is highlighted by the recent NICE Study, a multicentre study carried out in Japan using standardised methods 9. National statistics in Japan estimated COPD prevalence as 0.3%. Estimates based on spirometry in the NICE Study estimated COPD prevalence as 8.5% in those aged >40 yrs, almost a 30-fold difference. The NICE Study also highlights the value of using prevalence estimates to drive the awareness of COPD as an important public health problem.

Up to 2002, only 32 studies had been published on the distribution of COPD prevalence 3, which is an appallingly low number. Figure 1⇓ illustrates the chronological sequence of these studies and the prevalence estimates that they report. It would appear that there was a relative disinterest in COPD prevalence during the period of 1960–1990, followed by a steep increase in interest. It is also apparent that there is a wide divergence of prevalence estimates across the studies. Much of this can probably be attributed to different diagnostic and ascertainment methods used in the surveys.

Burden of obstructive lung disease

The lack of accurate population-based estimates of COPD prevalence in most countries prompted the formation of the BOLD initiative in 2002 by a group of investigators at the Kaiser Permanente Center for Health Research (Portland, OR, USA). BOLD is designed to provide strictly standardised, rigorous and practical methods for estimating the prevalence, social and economic burden of COPD. The data obtained from BOLD will enable governments and the private sector to make policy decisions on how to provide adequate and appropriate care for those suffering from COPD.

The primary objectives of BOLD are to: measure the prevalence of COPD and its risk factors in various countries around the world; estimate the burden of COPD in terms of its impact on quality of life, activity limitation, respiratory symptoms and use of healthcare services; and develop a validated model to project future burden of disease for COPD. BOLD also seeks to determine the extent to which variations in risk factors contribute to variations in the prevalence of COPD. Recognising the importance of standardising methods worldwide, BOLD worked collaboratively with PLATINO, an initiative of the Latin American Thoracic Society, to develop the methods. PLATINO subsequently used these methods to estimate COPD prevalence in five Latin American countries, i.e. Brazil, Mexico, Uruguay, Chile and Venezuela 22. The methods were piloted by BOLD in two countries: China and Turkey. Lessons learned from these pilots were incorporated into the BOLD methods after the pilots. The methods are now available for implementation worldwide, and BOLD is presently enrolling countries for implementation of the methods in 2005 and 2006.

The emphasis in BOLD is on building rigorous quality into the methods. An operations centre, located at the Kaiser Permanente Center for Health Research, supervises all aspects of the protocols. This includes sampling strategies, recruitment, translation of survey instruments, fieldwork, and data transfer and analysis. BOLD is designed primarily as a COPD prevalence survey among noninstitutionalised adults aged ≥40 yrs. This may take the form of, for example, a simple random sample, a stratified random sample, or some form of cluster sample. Countries must have their sampling plan reviewed and approved to ensure that the sample has good generalisability. Participating sites are expected to recruit a minimum of 300 males and 300 females in this age range. The proposed sample size of 600 individuals is designed to provide an acceptable level of precision for estimating prevalence at any given site.

The single most important outcome measure obtained as part of the BOLD protocol is spirometry before and after administration of an inhaled, short-acting bronchodilator. This is being carried out as the present diagnostic criteria for COPD use post-bronchodilator values for FEV₁ and the ratio of FEV₁/FVC%. Although standardised methods for performing spirometry are available and widely used, no single standard is universally applied. Proper training and ongoing quality control are essential in obtaining consistent high-quality measurements over time. The methods developed for BOLD meet or exceed the ATS standards for acceptable equipment and technique 23. An ongoing quality-control process that reviews every spirogram allows for early recognition of technical and technician-related problems, and pinpoints the nature of the deficiency, for example, difficulty in getting the participants to empty completely or blow hard enough. Therefore, the identified problems can be corrected quickly and field staff who are not capable of obtaining consistent high-quality spirometry can be deployed elsewhere. These methods were developed assuming that testing will often be done in the field, i.e. not in a climate-controlled pulmonary function laboratory. To optimise quality control in the BOLD study, sites are required to use the same spirometer. The spirometer used in the BOLD surveys (EasyOne^TM Spirometer; ndd Medizintechnik AG, Zurich, Switzerland) was selected as it provides a high degree of accuracy, robustness, portability and storage. Also, it can be used easily in the field and where there is no electric power available, since it operates on batteries. All spirograms for the BOLD survey are reviewed at a spirometry-reading centre in Salt Lake City (UT, USA) under the direction of R. Crapo and R. Jensen (Pulmonary Division and Department of Medicine, LDS Hospital and University of Utah, Salt Lake City, Utah, USA). Data that have been reviewed and graded are returned to the country site so that they can be used for field ongoing training of the staff.

In addition to spirometry, study participants are administered a standardised questionnaire covering respiratory symptoms, health status, activity limitation and exposure to potential risk factors, such as tobacco smoke, occupational risk factors and biomass exposure. All questionnaires are forward- and back-translated, again using standardised methods to ensure comparability of the questionnaires in all countries.

All data from the field sites are transmitted through secure, encrypted Internet transfer to the operations centre. Use of the same type of spirometer and software throughout the project makes training more efficient, and data compilation and transfer easier and less susceptible to mistakes. BOLD uses a “train the trainer” approach to training country staff. This entails having the key staff in each country attend a 5-day training programme, which includes all aspects of the BOLD protocol. The trainers then have the responsibility of training their field teams. Quality-control surveillance, especially of the primary outcomes, allows ongoing monitoring of the field staff.

An additional part of the BOLD initiative is the development of an interactive, web-based model to estimate the economic burden of COPD. This model, once completed, will be accessible to all on the BOLD website, so that country-specific estimates of the prevalence and economic burden of COPD can be developed. Preliminary data from BOLD in the PLATINO countries 24 and in the Chinese and Turkish pilot studies are of great interest. BOLD is now entering phase 3, in which interested countries can apply to participate (www.boldcopd.org).

In conclusion, COPD is a huge and growing burden worldwide as a result of the changing demographics of the populations in developing countries and the tobacco epidemic. Better descriptive data on the prevalence, morbidity, mortality, and social and economic burden of COPD are urgently needed in order to focus the attention of the healthcare community and planners on this growing problem. Attention must also be paid to the development and acceptance of standardised methods that can be used worldwide to accumulate these data. Interventions that are appropriate for each country should be developed so that the trend can be reversed and COPD can, ultimately, be prevented and treated.

Summary

COPD was the sixth most common cause of death worldwide in 1990 and is projected to become the third most common cause by the year 2020. The global increase in deaths from chronic lung disease has provoked discussion among COPD researchers about the adequacy of current structures for monitoring the disease burden from COPD in the absence of population-based registries. In the current paper, selected insights gained from cancer surveillance over the last half-century that may be relevant to COPD surveillance are discussed. In particular, the current authors consider the goals of surveillance of noncommunicable disease and describe examples of how routinely collected mortality data from death certificates or incidence data from population-based registries can be used to monitor temporal trends, identify high-risk geographical or demographic subgroups, and focus disease-control activities. Several barriers that may impede efforts to establish population-based registries for COPD are identified. Finally, approaches that may strengthen ongoing efforts to monitor and reduce the disease burden from COPD are suggested.

Introduction

COPD was the sixth most common cause of death worldwide in 1990, but is projected to become the third most common cause by the year 2020 25. COPD is already ranked fourth as a cause of death in economically developed countries 26. COPD and diabetes mellitus are the only two common conditions for which the age-standardised death rate increased in the USA during the 1990s 27. Mortality from COPD continues to increase in many other developed and developing countries due to ageing of the population and the global spread of cigarette smoking 28. Appropriate labelling of undiagnosed COPD patients or at the time of death, and more recognition of the disease worldwide will probably contribute towards increased COPD awareness.

The increase in deaths from chronic lung disease has raised questions about the adequacy of surveillance data currently available to monitor COPD, given the importance of the disease. This discussion will include lessons learned from cancer surveillance over the past half-century and how insights from cancer surveillance may help to inform efforts to improve surveillance of COPD. To address this, members of the American Cancer Society (ACS) will first consider the overall goals of disease surveillance, as they relate to both communicable and noncommunicable diseases. Examples will then be provided of how mortality data collected from death certificates, and incidence data from population-based registries can be used to monitor temporal trends, identify high-risk geographical or demographic subgroups, and focus disease-control activities. Several challenges or barriers that may impede efforts to establish population-based registries for COPD are identified. Finally, approaches that may help to strengthen future surveillance efforts are suggested.

Goals of disease surveillance

The Dictionary of Epidemiology defines disease surveillance as “the continuing scrutiny of all aspects of occurrence and spread of disease that are pertinent to effective control” 29. An essential feature of such activities is the application of information gained from surveillance activities to strengthen disease-control efforts. While the methods and terminology that apply to surveillance of noncommunicable diseases (e.g. cancer or COPD) differ somewhat from those used to monitor infectious diseases, such as malaria or HIV infection, the fundamental purpose is the same. The goal is to measure the burden of disease in well-defined populations (in terms of incidence, prevalence, survival, disability, mortality, economic costs, etc.) and then to use this information to improve disease control.

Examples of use of mortality data

Routinely collected information from death certificates has been useful in monitoring temporal trends in age-standardised death rates from cancer (all sites combined, or specific cancer sites) and for identifying geographical areas and demographic subgroups at a particularly high risk. Mortality data have been collected across the USA since 1930; considerable attention has been devoted to establishing systematic rules for coding cancer as the underlying or contributing cause of death to allow temporal and geographical comparisons. One use of mortality data is illustrated in figure 2⇓, showing temporal trends in the age-standardised death rate (per 100,000 per year) from selected cancers among males and females in the USA from 1930 to 2001, the most recent year for which data are available. This figure illustrates that lung cancer has dominated cancer death rates among males in the USA since 1950, and among females since the 1980s. It also depicts the decrease in the age-standardised death rate that has occurred since 1990 for cancers of the lung, prostate and colorectum in males, and cancers of the breast and colorectum in females 30, 31. This figure has been updated annually by staff at the ACS since the 1960s. When combined with the temporal trend in per capita cigarette consumption, it provides an impressive visual display of how the lung cancer epidemic parallels the rise and fall of cigarette consumption in the USA, as it does in many other countries.

In certain respects, the temporal trend in cancer mortality rates can provide a more reliable indication of the actual trend in disease occurrence than is provided by incidence data. Death rates are less susceptible to fluctuate with the introduction of new diagnostic or screening tests than incidence rates. This point can be illustrated by comparing the trend in incidence with that in mortality for all cancers combined in the USA during the early 1990s 31. A large increase in the incidence rate of all cancers combined from 1973 to 2001 is seen among males in the early 1990s, followed by an equally abrupt decrease. The increase coincided with the widespread introduction of prostate-specific antigen (PSA) testing, and the decrease occurred when PSA testing reached equilibrium in the population. The introduction of PSA testing had minimal impact on the death rate from prostate cancer or all cancers combined, and no effect on the incidence or death rate from cancer in females. However, its effect on cancer incidence in males provides a remarkable illustration of how the introduction of a new, sensitive method of disease detection can distort the apparent temporal trend in incidence in a population where subclinical disease is common.

The geographical distribution of mortality from certain cancer sites can be used to identify high-risk subgroups and to target intervention efforts. For example, the death rates from cervical cancer have been mapped by state economic area among White females in the USA from 1970 to 1994 32. Areas with the highest death rates from cervical cancer are concentrated in Appalachia and in Southwest Texas, where widespread poverty and cultural factors limit access to effective screening tests, such as Papanicolaou (Pap) testing. Evidence that mortality from cervical cancer remained high in large areas of the USA, coupled with survey data demonstrating low utilisation of Pap testing among females in lower socio-economic groups, helped to motivate the Breast and Cervical Cancer Early Detection Program. This initiative provides mammography and Pap testing to females who would otherwise not have access to standard screening tests.

Another example of how geographical patterns in cancer mortality can stimulate potentially important hypotheses is illustrated by a map contrasting mortality from lung cancer with that from COPD. Figure 3⇓ shows that the age-standardised death rate from lung cancer among White males during the period of 1970–1994 was highest in the tobacco-growing states of southeastern USA 32, 33. In contrast, areas with the highest mortality from COPD among males in the USA are concentrated in Western mountain states (fig. 4⇓) 33. A similar difference in the geographical patterns is seen among females. Figure 5⇓ shows that the highest lung cancer death rates among USA females are on the East and West coasts (as well as in Nevada), whereas the areas with highest COPD mortality include Colorado, Wyoming and other mountainous areas (fig. 6⇓) 33.

The current authors considered several factors that might contribute to the divergent geographical patterns in mortality from lung cancer and COPD, given that tobacco smoking is a major cause of both diseases. Migration patterns among patients who develop COPD seem unlikely to explain the observed pattern. However, immigration of ethnic minorities to different parts of the USA might also affect the rates observed, as some races might be more prone to lung cancer, or COPD, than others and time of migration might impact upon risks of cancer. Occupational exposures to cofactors, such as silica from hard rock mining, or grain dust from farming might possibly contribute to the geographical pattern in males, if not females. A more plausible alternative explanation may be that pulmonary dysfunction is more likely to be recognised in regions with higher altitude and lower ambient oxygen concentration.

Descriptive analyses of temporal or geographical patterns of disease are useful in generating potentially valuable hypotheses, but not in determining the reason(s) for such patterns. More definitive answers can be obtained only from other study designs. However, descriptive analyses can reflect fundamental characteristics of underlying disease processes. For example, figure 7⇓ compares age-specific death rates from lung cancer and COPD among males and females in the USA from 1997 to 2001 30. The death rates from both diseases increase markedly with age; however, the increase at older ages is even greater for COPD than for lung cancer. A decrease in the death rate from lung cancer is seen at older ages, and is believed to reflect incomplete diagnosis of lung cancer in the elderly and lower death rates among birth cohorts who preceded the era of maximum cigarette smoking in the USA 34. In contrast, the death rate from COPD continues to increase with age. Progressive loss of lung function occurs with ageing. The development of symptoms in the elderly may lead to pulmonary function testing and documentation that an individual meets the functional criteria that define COPD. It is interesting, in light of the geographical difference between COPD and lung cancer mortality discussed previously, that the death rates from these two diseases are more strongly correlated in persons aged <60 yrs than in those >60 yrs. The correlation coefficient between the death rate from lung cancer and that from COPD, during the time period 1997–2001 across all 50 USA states, was higher among males aged 30–59 yrs (r = 0.61; p = 0.002) than among males aged ≥60 yrs (r = 0.42; p<0.001) 27. This suggests that at least some of the geographical differences in mortality from lung cancer and COPD may reflect factors that influence diagnosis at older ages.

Value of incidence registries

Population-based cancer incidence registries are used to monitor the number and rate of new diagnoses, to examine histological and other subtypes of disease that cannot be identified on death certificates, and to collect information on other parameters, such as stage at diagnosis, tumour grade, and absolute and relative survival rates. The National Cancer Institute (NCI) supports a network of tumour registries called the Surveillance, Epidemiology, and End Results (SEER) Program 35. SEER registries were established in Connecticut, Hawaii, Iowa, New Mexico, Utah, and the metropolitan areas of Detroit and San Francisco-Oakland as early as 1973. The SEER network has been expanded, subsequently, to include the 13 counties of Seattle-Puget Sound, the metropolitan area of Atlanta and 10 rural Georgia counties, Native American populations in Arizona and Alaska, and all of California, Kentucky, Louisiana and New Jersey 35. More recently, many other states, not included in the SEER Program, have developed state-wide cancer registries through the National Program of Cancer Registries, supported by the Centers for Disease Control (CDC) 36. Most of these registries are too new to examine long-term trends, but do provide information about cancer incidence rates for more than half of the USA population 37.

Long-term trends in cancer incidence, as measured in the original SEER nine-state areas, have been used to monitor changes in the incidence of specific cancer sites and histological subtypes 38. For example, between 1974 and 1994, the incidence of adenocarcinoma of the oesophagus increased in both White and Black males in the USA, whereas the incidence of squamous cell carcinoma of the oesophagus decreased. Among White males, the incidence rate of these two histological subtypes actually crossed over in the late 1980s, so that adenocarcinoma became the predominant histological subtype. The rise in oesophageal adenocarcinoma observed here, and in many other industrialised countries, coincides with a major increase in obesity 39, 40.

SEER data have also been used to project the future increase in the number of cancer cases that can be expected, due to growth and ageing of the population 41. It has been projected that, if the age-specific incidence rate from all cancers combined were to remain constant from 2000 through to 2050, and the population was to grow and age according to projections by the USA Census Bureau, then the total number of cancer cases in the USA would double between 2000 and 2050 41. The increase would be even larger in proportionate terms for COPD than for cancer, because the population growth is greatest in the oldest age groups where COPD risk outstrips the risk of lung cancer (fig. 7⇓).

The information on cancer incidence collected by state tumour registries can also be used to identify high-risk areas where special intervention efforts may be required. This is particularly relevant to cancer sites for which early detection improves prognosis. For example, Roche et al. 42 used information from the New Jersey State cancer registry to identify two areas within the northeastern corner of the state where an unusually high percentage of breast cancers were classified as distant-stage disease at the time of diagnosis. Contrary to expectation, the high-risk areas were surrounded by the highest concentration of mammographic facilities in the state. However, females in the high-risk areas were more likely to be foreign-born, Hispanic or African American, have lower educational and socio-economic status, and to be less proficient in English than the average citizen of New Jersey. These results suggested that poverty and barriers of language and/or culture were the probable causes of this disparity. The information contributed to the development of a Comprehensive Cancer Control Plan for New Jersey, and led to targeted educational services to overcome these barriers.

Value of collaboration and coordination

An important development in cancer surveillance in the USA is the effort to coordinate surveillance activities across multiple organisations and agencies. For the past 6 yrs, the major organisations involved in cancer surveillance have published an annual report to the nation on the status of cancer 41, 43–47. Organisations that collaborate on this effort include the ACS, CDC, NCI and the National Association of Central Cancer Registries. Another forum that encourages collaboration is the National Coordinating Council for Cancer Surveillance, which includes representatives from all of the previously mentioned organisations plus the American College of Surgeons, the American Association of Central Cancer Registrars and the Armed Forces Institute of Pathology. Its mission is to coordinate cancer surveillance activities within the USA, ensure that scarce resources are used optimally, and address issues of common concern 48.

Expanding definition of cancer surveillance

Increasingly, the framework for cancer surveillance is expanding to encompass a continuum of factors from health and primary prevention, through screening, diagnosis, treatment, quality of life, palliative care and end of life 48. This expanded definition not only includes population-based data on individual risk factors, such as tobacco use, obesity or poverty, but it also includes social and legislative policies, such as cigarette taxes, restrictions on smoking in restaurants and workplaces, community initiatives to increase physical activity and federally funded nutrition programmes. The data being collected on tobacco are relevant to COPD as well as cancer. The information currently being collected on cancer treatment, comorbidity and quality of life is far more limited. Substantial resources and/or new technologies will be required to collect valid, population-based data on these issues.

Challenges of COPD surveillance

Several barriers must be addressed and overcome in order to improve the surveillance of COPD. One of the most important is to establish a uniform definition of COPD so that surveillance data can be compared across different regions of the world. In the past, the criteria used by the ATS 49 differed slightly from those of the ERS 50 and the WHO's GOLD 4. Even if a single definition is agreed upon, the extent of pulmonary function testing in the population will influence the completeness of diagnosis and the interpretation of temporal trends. Furthermore, the reliance on functional rather than anatomical or pathological criteria will cause the incidence of newly diagnosed disease to fluctuate with cofactors such as infectious or environmental exposures that impair pulmonary function acutely and affect the likelihood that COPD will be diagnosed.

Another consideration that may limit the value of population-based incidence registries for COPD is that chronic lung disease is more difficult to categorise into distinctive and mutually exclusive subtypes of disease than cancer. The same patient may exhibit characteristics of both chronic bronchitis and emphysema, even though these conditions may represent different pathological processes. Furthermore, the perception that COPD is a degenerative disease that predominantly affects the elderly may weaken public support for resource-intensive registries.

Recommendations

The next step in efforts to improve the surveillance of COPD might be to convene a working group of pulmonary epidemiologists to characterise the existing data resources and to consider the opportunities and challenges more systematically. This process might stimulate greater use of existing data. It could encourage collaboration, help to define the priorities in COPD surveillance, and clarify the rationale for new initiatives. There may be opportunities to extend research approaches that have been used for other diseases to COPD. For example, it may be possible to using existing national registries in Scandinavia to measure population-based incidence rates, or to conduct special studies comparing high-risk populations (such as in certain regions of China) to lower risk populations who have migrated from China to other countries. Well-designed, collaborative research projects could attract additional resources for the research, prevention and treatment of this important cause of human suffering.

Summary

In addition to understanding the epidemiology of a disease, estimates of the disease costs are important for appreciating the overall disease burden. Estimates of the economic burden of a disease are important for informing policy decisions. Health economic models can be used to estimate the future burden of disease or to understand the value of an intervention. Importantly, the cost estimates of the components of a disease are vital for developing health economic models of diseases. The focus of this section of the paper is to review the economic burden of COPD and COPD disease-state models. Published cost estimates from Spain, Sweden, the USA, the Netherlands and Italy are reviewed. Additional attention is given to the cost of exacerbations, which is an important component of the overall cost of COPD. Finally, the use of health-state models in COPD is briefly discussed and the BOLD study model is introduced.

Introduction

In addition to understanding the epidemiology of a disease, estimates of the disease costs are important for appreciating the overall disease burden. Estimates of the economic burden of a disease are important for informing policy decisions. Healthcare decision makers use information on the magnitude of costs associated with a disease and what might reasonably be expected in the future when making resource-allocation decisions. Estimates of the components of the overall healthcare costs of a disease can help decision makers target interventions where they may have the most impact on overall disease-related healthcare costs because the component is a driver of the overall burden of the disease.

The cost estimates of the components of a disease are vital for developing health economic models of diseases. Disease-state models can be used to estimate the economic burden of a disease where only certain pieces of information may be available, in addition to estimating the value of new interventions in the disease. Thus, disease-state models can serve as valuable tools for decision makers when setting policies.

In COPD, the burden of the disease has received increased attention in recent years. Population-based and individual patient costs have been reported from several populations. In addition, COPD disease-state models have been published in the literature. The focus of this section is to review the economic burden of COPD and COPD disease-state models. This section is divided into two parts as follows. The first part focuses on estimates of the cost of COPD, and methods for estimating disease costs are reviewed and estimates of COPD costs from throughout the world are summarised. In the second part, the COPD disease-state models are briefly reviewed and the BOLD health economic model is introduced.

Cost of disease

Economic cost studies of diseases are aimed at quantifying some of the effects that a disease has on both the patients themselves and society. This method has been widely used during recent years. However, in order to analyse and comprehend a pharmaco-economic study, it is necessary to focus on some essential aspects 51 as follows: 1) whether the study is based on the prevalence or incidence of the disease; 2) whether data collection will be from the “top-down” or from the “bottom-up”; and 3) how the direct and indirect costs will be defined, calculated and considered.

Modelling in COPD

Health economic models have been used in several chronic diseases to evaluate the costs of a disease, as well as the value of new interventions 68–72. They can serve as valuable tools for decision makers when making resource-allocation choices. Health-state models attempt to capture the nature of the disease in a mathematical form, in order to answer questions about the costs of the disease. In COPD, health-state models have been used to estimate the burden of disease and to examine the value of interventions.

Summary

Tobacco smoking is undoubtedly the most important risk factor for COPD. Smoking causes COPD more commonly than had been previously recognised, with up to 50% of long-term smokers developing the disease. However, smoking accounts for a smaller proportion of cases of COPD than previously realised, with a population attributable risk for COPD with smoking of ∼50%. It appears that it is the interaction with other risk factors that determines an individual smoker's susceptibility to COPD. In addition to public health and political measures to reduce the prevalence of smoking, the investigation and modification of these other risk factors is essential if the global burden of this disease is to be reduced.

Introduction

In 1964, the USA Surgeon General's report warned that tobacco smoke was the most important cause of chronic bronchitis. The statement was expanded in the next report published in 1967, implicating tobacco in the aetiology of emphysema 79. Since then, thousands of articles have been devoted to establishing tobacco as the leading risk factor in the development of COPD. Despite the wealth of evidence available, the exact mechanisms by which tobacco causes COPD are not yet fully understood and there remain important unanswered questions. Given the strong association, why do only 15% of smokers develop COPD, or is the incidence actually higher? What are the causes of the individual variation in response to tobacco smoke seen in smokers? What other factors play a role in the pathogenesis of COPD and are they amenable to modification?

An individual's susceptibility to COPD is determined by the interaction between “host factors” (i.e. genetics, airway hyperresponsiveness, lung growth, sex and race) and exposure to “environmental factors” (i.e. tobacco smoke, passive smoke, marijuana smoke, pollution, occupational dusts/chemicals, socio-economic status, respiratory infections and diet). In this section, the role of host and environmental factors implicated in the pathogenesis of COPD is briefly reviewed.

Genetic risk factors

There is a significant body of evidence that supports a hereditary component to the development of COPD. Clustering of COPD cases has been observed in families, and several studies have demonstrated an increased incidence of COPD in relatives of cases when compared to controls. Concordance in lung function impairment has been witnessed in monozygotic twins, but not replicated in studies on dizygotic twins 80.

Numerous genetic abnormalities have been proposed to contribute to the pathogenesis of COPD (e.g. α₁-antitrypsin, α₁-antichymotrypsin, cystic fibrosis transmembrane regulator, vitamin D-binding protein, α₂-macroglobulin, cytochrome P450 A1, blood group antigens, human leukocyte antigen locus and immunoglobulin deficiency). Their importance lies not only with better understanding of the mechanisms by which COPD develops, but also in identifying novel therapeutic targets for the primary or secondary prevention of COPD.

The most extensively researched of these genes is α₁-antitrypsin, a potent antiprotease, which also inhibits leukocyte elastase. Over 70 variants have been identified, of which the most common are M, S and Z with allele frequencies of 0.93, 0.05 and 0.02, respectively. In addition, there are mutations that affect the function of α₁-antitrypsin. The severity of the deficiency is determined by the genotype inherited, with the ZZ genotype having the most severe deficiency, with α₁-antitrypsin levels being ∼15% of normal 80. Homozygotes of the Z variant have a significantly accelerated decline in lung function at a young age, with a synergistic interaction with exposure to cigarette smoking. However, the overall contribution of this genotype to the global COPD burden is small 80.

Airway hyperresponsiveness, asthma and atopy

The concept that airway hyperresponsiveness may be a host factor that predisposes individuals to COPD is known as the “Dutch Hypothesis” 81. It has been proposed that asthma, emphysema and chronic bronchitis are different manifestations of the same underlying disease processes and that the actual presentation depends on host factors such as age, sex, atopy and the relative severity of airway hyperresponsiveness, which modulate the response to environmental factors such as cigarette smoke. Clinical studies have also demonstrated an association between bronchial hyperresponsiveness and COPD. What is not yet clear is whether the hyperresponsiveness seen in COPD is a manifestation of the airways narrowing or a causative factor in its pathogenesis.

The differentiation between asthma and COPD can be difficult in older adults, since a component of irreversible airflow obstruction and a reduced diffusion capacity may be features of both conditions 81, 82. Amongst adults with obstructive lung disease, those with a diagnosis of asthma represent the largest subgroup 83. Furthermore, longitudinal cohort studies have shown that asthma in adults aged in their late 40s is strongly associated with symptoms of chronic bronchitis, a diagnosis of emphysema, or fulfilling COPD criteria 20 yrs later. The risks are substantial, over 10-fold, when compared with nonasthmatics, even after adjusting for confounders such as smoking 84.

Similarly, there is evidence to suggest that atopy (as indicated by a raised blood eosinophil count) may have a role in the pathogenesis of COPD, independent of its association with bronchial hyperresponsiveness 81.

Lung growth in utero and during infancy

There is evidence that significant insults during the vital period of lung growth in utero (after 16 weeks) may have implications for lung function in adulthood 85. Insults include maternal smoking, which has been shown to have several adverse effects on lung development and poor foetal nutrition. Premature babies with low birthweights have been shown to have impairment of FEV₁ that persists into late childhood. More importantly, birthweight is correlated with FEV₁ in adults and the death rate from chronic bronchitis has been shown to be inversely proportional to birthweight. Similar relationships also exist with weight at 1 yr of age, when the lungs are still developing.

Sex

The subject of sex differences in COPD is controversial, with current evidence suggesting differential sex risks for COPD in different communities. Among the possible explanations for observed sex differences are differences in lung morphology, smoking, hormonal factors, differences in inflammatory response between the sexes, and occupational confounders 86.

Tobacco smoke

Tobacco smoke is the only environmental risk factor whose contribution to COPD is undisputed, satisfying all of the Bradford Hill criteria for proving causation. Although the link between cigarette smoking and COPD is well established, there remain some areas of uncertainty in the relationship, including the basis of the wide range of susceptibility to cigarette smoke within and between populations.

The effects of tobacco smoke on the lung are multiple, including the release of elastase from inflammatory cells in the lung, oxidative stress and inactivation of α₁-antitrypsin. Of particular importance is the effect whereby tobacco smoke increases intrapulmonary elastase activity, inactivates intrapulmonary elastase inhibitors and blunts the repair of the injured pulmonary extracellular matrix 87. Tobacco smoke is also a rich source of oxidants and upregulates pathways that decrease antioxidant capacities within the lung 88. Through the recruitment and stimulation of alveolar macrophages and neutrophils, an acute or chronic inflammatory process is established within the lung, which leads to further tissue damage 89.

While it is generally considered that only 15–20% of smokers develop COPD, recent evidence indicates that the figure may be closer to 50% 90. This study also suggested that the population attributable risk for COPD with smoking is ∼50% rather than the 90% currently proposed. These observations suggest that COPD occurs more commonly in smokers, but is responsible for a smaller proportion of cases of COPD than previously realised. The importance of stopping smoking to reduce the progression of the disease has been well demonstrated (fig. 10⇓) 90, 91.

Smoking remains the major cause of mortality from COPD, which has been increasing over the past two decades, in contrast with a decreasing mortality for other major chronic diseases (fig. 11⇓) 92. In 2000, it has been estimated that there were ∼1,000,000 premature deaths in the world attributable to smoking 93.

Environmental tobacco smoke

Notwithstanding the limitations of studies investigating the association between passive smoking and lung disease, there is evidence that maternal smoking in utero is associated with impaired lung development and reduced lung function in adults 94. Children living in houses in which one or more of the parents smoke have more respiratory tract infections. There is also evidence in adults that exposure to second-hand smoke is associated with reduced lung function, including impaired diffusing capacity, indicating the presence of emphysema 95. The effects of public health measures limiting smoking in the workplace and public areas are awaited with interest.

Marijuana smoke

Compared to tobacco, marijuana has been somewhat neglected by researchers investigating the aetiology of COPD. This is surprising, considering that marijuana is the most commonly used illegal drug worldwide, is constitutionally similar to tobacco (aside from the two active ingredients), is smoked without filters, and results in a four-fold greater particle burden on the respiratory tract than tobacco smoking 96. Conversely, fewer joints are smoked by marijuana smokers than cigarettes by tobacco smokers, and, while tobacco smoking is often a lifelong habit, most marijuana smokers cease smoking in later adult life.

Since a significant number of marijuana smokers also smoke tobacco, future research needs to be directed at investigating the possibility of a synergistic relationship between the two substances. In addition, a dose equivalence with tobacco would be informative in assessing the relative risk of COPD for marijuana smokers.

Air pollution

Historically, the association between respiratory disease and pollution became apparent during the 1950s when the great “British Smogs” caused an acute mortality epidemic from chronic bronchitis and respiratory failure 97. Subsequent studies from other countries confirmed these findings, demonstrating the adverse effects of air pollution on lung function in childhood and exacerbations of COPD resulting in hospital admissions and sick days due to respiratory symptoms 97, 98. The magnitude of the role of outdoor air pollution in the pathogenesis of COPD is variable by country and geography.

Although air-pollutant emissions are dominated by outdoor sources, human exposures are a function of the level of pollution in places where people spend most of their time. Human exposure to air pollution is dominated by the indoor environment. Globally, the largest source of indoor pollution comes from cooking and heating with solid fuels, such as dung, wood, agricultural residues and coal. These fuels emit substantial amounts of pollutants. It has been estimated that indoor smoke from solid fuels causes ∼20% of cases of COPD worldwide 99.

Occupational dusts/chemicals

Occupational exposure to dusts, chemicals and fumes is a risk factor in the development of COPD 100. The agents particularly implicated with impaired lung function are silica, coal dust, fibreglass, sawdust, freons and solvents. There is a relationship between the degree of lung function impairment and the intensity and duration of exposure, which supports causality. It is likely that the interaction between occupational exposure and other risk factors determines whether an individual develops COPD 101.

Intriguingly, it has been shown that males with significant occupational exposure have greater lung function impairment if they also have a positive airway challenge test than those who do not 100. This suggests that occupational exposure constitutes a risk factor in those who have an underlying predisposition to airways disease or, alternatively, that airway hyperresponsiveness represents a component of the airway response to injury.

Given that 10–20% of cases of COPD may be attributable to occupational exposures, occupational and public-health policy makers and clinicians need to address this potential avenue of disease causation and prevention 102, 103.

Socio-economic status

COPD is linked to socio-economic status, measured by income and educational level, in terms of severe exacerbations, prevalence and mortality. While a greater proportion of people in the lower socio-economic group smoke, this is insufficient to explain the association. The strength of the association is marked with the difference between the lowest and the highest socio-economic group equivalent to the deterioration in lung function seen in 10 yrs of ageing in nonsmokers. The reasons for the association are multiple, but may include poorer housing, poorer nutrition, use of fossil fuels without adequate ventilation, and greater frequency and severity of respiratory tract infections 104.

Respiratory infections

There is evidence to suggest that an insult inflicted on the lungs by infective agents during childhood, when the lungs are still developing, may cause permanent damage that predisposes to COPD in later life. This hypothesis arose following the observation in Great Britain that regional rates of chronic bronchitis in adults were similar to the regional rates for respiratory illness in children 85. Subsequent studies reported that bronchiolitis, croup and pneumonia in early childhood may cause long-term functional impairment 105, 106. The impairment may be substantial with a mean loss in FEV₁ of 0.65 L reported in males who had pneumonia before the age of 2 yrs, a reduction in FEV₁ about twice that associated with lifelong smoking 105. Since follow-up commences after the childhood infection in many studies, it is not clear which component of functional impairment was a result of the infection and which was due to a predisposing factor that led to the infection in the first place.

There is also evidence to suggest that latent adenoviral infection or colonisation of the airways with bacterial pathogens such as Haemophilus influenzae or Branhamella catarrhalis, or other pathogens such as Chlamydia pneumoniae may contribute to the pathogenesis of COPD 107–109. The role of ongoing infection in the lower airways is also suggested by the recent findings of coincident bronchiectasis in up to half of subjects with moderately severe COPD 105.

Diet

The important contribution of oxidative stress to lung injury in COPD was the foundation for considering diet as a risk factor for the disease 88. This led to the theory that the antioxidant properties of certain nutrients, such as vitamin C, beta carotene, selenium and copper, may modulate an individual's susceptibility to oxidative damage. It has also been proposed that a diet rich in omega-3 fatty acids may inhibit arachidonic acid production, thereby protecting against bronchoconstriction. While it has been demonstrated that a diet rich in fruit and vegetables is associated with a reduced risk of COPD 110, the effect that modification of diet has on the prevalence of COPD is yet to be determined.

In conclusion, extensive epidemiological investigation has shown that tobacco smoke is the single biggest risk factor for chronic obstructive pulmonary disease. However, it appears that it is the interaction with other risk factors that determines an individual smoker's susceptibility to chronic obstructive pulmonary disease. In addition to public health and political measures to reduce the prevalence of smoking, investigation and modification of these other risk factors is essential if the global burden of this disease is to be reduced.