To the best of our knowledge, this is the first study to incorporate the effects of age-specific time trends on COPD mortality separately for men and women when estimating future trends in COPD mortality. Our findings suggest that overall COPD mortality trends will continue to decline, but that this decline will be later for women than for men.
As our focus was on comparing the COPD mortality rates of men and women, we did not compare our overall COPD mortality trends with those found in other studies directly. However, our findings on the decline of COPD mortality rates among men are in agreement with studies in many other countries. Using an age adjusted trend analysis, M.J. Goldacre and colleagues [11] demonstrated an overall decline in COPD mortality for men in England over the period 1979-1998. When stratified by age, they also found a decline for all age groups. An earlier study of COPD mortality trends in Canada for the period 1951-1995 [12] reported a decline in the overall COPD mortality for men, with a decline in the 65-74 year age group and an increase among men over 75 years of age. In the United States (USA), COPD mortality rates increased from 1968 to mid 1980s, [13] stabilized, and then declined after 2000 [14]. Similarly, the mortality trends for men in France began to decline steadily after 1985, and further declined between 2000 and 2002 [15].
Our findings on increasing COPD mortality trends for women pre-1997 parallel the patterns observed in studies in other industrialized countries. However, our findings on the decline over recent years are not in agreement with these studies, as recent data are not yet available for comparison. In the United Kingdom (UK), rates for women continued to increase between 1978 and 1998, with the greatest increase being in women over 65 years of age. This finding was consistent across the different combinations used to code the underlying cause of death [11]. We do not have more recent UK data for a comparison of trends beyond 1998. Similarly, COPD mortality rates for women in Canada continued to increase from the late 1950s to 1995, with an increase in age-specific mortality rates in women aged 65 years and over [12]. More recent data from the United States suggest that COPD mortality rates among women tripled between 1980 and 2000, and in 2000 were higher than the rates in men [14]. Our findings are in agreement with a study from France (where recent data were available) which found that female mortality rates increased slightly between 1979 and 1999 and declined thereafter [15].
The difference in mortality trends between men and women may be explained by a number of factors. The main causal factor is tobacco exposure. COPD mortality trends in men first began to rise during the 1950s, with the maximum death rates for the age group 80-84 in 1980. The trends in COPD mortality rates among men mirrored the pattern of the uptake of smoking 50 years earlier. By 1945, almost 72% of men were smokers. Smoking rates among men then fell sharply during the 1950s and early 1960s when concerns about effects on health were first raised [16, 17], but, as a result of the aggressive marketing of tobacco products to younger smokers in the 1960s and 1970s, subsequently remained stable until the early 1980s and the advent of organized Quit campaigns.
Similarly, the patterns of COPD mortality trends among women have reflected the patterns of smoking uptake among women. In the years leading up to and during World War II, the advent of movies depicting glamorous images of women smoking, the entry of women into the workforce, and the development of cigarette brands which were attractive to women resulted in widespread uptake, with more than one in every four women reported as smoking by 1945 [18].
While many women also quit in response to the health concerns raised in the later 1950s and early 1960s, the effects of this were offset by increased uptake by women who were born in the 1950s and 1960s, corresponding to aggressive marketing by the tobacco companies, and the increased economic and social freedom of women since that time. Smoking rates therefore barely changed until the late 1980s, but have been declining steadily since 1998 [19].
In addition, these effects may be prolonged by the action of other risk factors such as genetics, indoor and outdoor environment, and lifestyle [20, 21]. Other factors include the possible biological and physiological differences between men and women [22]. Women seem to be less likely to be diagnosed with COPD (even with severe disability), and more likely to be misdiagnosed with asthma than men. Chapman and colleagues [23] found that physicians were less likely to diagnose COPD in women (49%) than in men (64.6%), where both presented with hypothetical cases of cough and dyspnea. Women may also differ from men in their response to treatment and disease management [24]. All of these factors subsequently diminish the quality of life with a longer survival time.
Comprehensive tobacco control policies and the decline in tobacco consumption since the early 1970s may explain a considerable component of the declining trend of COPD mortality in women in Australia. As is discussed above, this trend has not been observed in many other countries with a similar population-mix. This may reflect the earlier decline in tobacco consumption in Australia, compared to countries such as the UK and the US; and the considerable success of Australia's QUIT Campaign, compared to smoking cessation strategies in other countries.
There are few published forecasts of age-specific COPD mortality. Two studies from the Netherlands [20, 25] developed a dynamic multistate life table model for modelling the prevalence, mortality and health care costs of COPD. Their simulations suggested an increase in COPD-related deaths for both men and women by 2025. These increases became more pronounced in women than men when estimates of smoking prevalence were factored into the model. It is difficult to compare our forecasts with the forecasts from this study because of the differences in modeling approaches and assumptions. For example, these studies assumed constant age- and sex-specific mortality rates for each of disease severity and smoking class. In contrast, our modeling approach produces forecasts based on the entire functional form of the age-mortality curve over time (although they do not take the intervention effects of smoking behavior into account).
The modeling and forecasting approach adopted in this study has a number of strengths. It has the ability to model the functional form of age-related changes in mortality rates over time and make predictions for specific age groups. This enables us to understand the different patterns of disease progression for men and women in younger, middle and older age groups. It provides important insights into the natural history of COPD, which has implications for policy and program development for at-risk age groups.
A second major strength of our approach is the improved modeling of sudden directional changes or jumps in the trends of an outcome. An earlier study of COPD deaths in Australia between 1993 and 2003, and one which generated a considerable amount of debate, modeled the data using linear regression with autocorrelated errors, but failed to capture the sudden change in COPD mortality rates around 1997 [26]. While it is important to understand the reason for this change in the disease pattern, it is equally important for the models and the predictions which they generate to allow for this structural change. The failure to do this is a common phenomenon in trend studies of disease outcomes such as prevalence, incidence and mortality.
Some limitations should be considered when interpreting the results. Our models and the subsequent forecasts do not take into account trends in tobacco consumption, which remains the most important causative factor for COPD in Australia, or the fact that there is a considerable "lag time" between consumption and COPD mortality. The models also fail to take into account intervention effects such as the commencement of smoking cessation programs. There are also well-recognized limitations of mortality data which we do not incorporate into the models. The ascertainment of COPD is often incomplete, as the diagnosis may not have been made during life [27], or may simply have been overlooked in completing the death certificate. COPD is listed as the underlying cause of death only a minority of the time [28]. There have also been changes in the coding of respiratory deaths in the ICD between the third revision in 1922 and the tenth in 1999. These changes confound any interpretation of long-term trends. At present our modeling techniques do not allow for birth-cohort trends, although our diagnostic tests revealed little evidence of cohort trends, with a negligible impact on the estimated basis functions and forecasts.
The national guidelines for COPD management emphasize the cessation of smoking as the only intervention proven to slow the decline in lung function [29, 30]. From the graphs it seems that mortality is declining at a slower rate for older women. This is an important finding because this is probably the result of a cumulative exposure to smoking, passive smoking and other lifestyle variables.
