Epidemiological modelling of routine use of low dose aspirin for the p
http://www.100md.com
《英国医生杂志》
1 Discipline of General Practice, School of Medicine, University of Tasmania, Hobart, TAS 7001, Australia, 2 Department of Epidemiology and Preventive Medicine, Monash University, Alfred Hospital, Melbourne, VIC 3004, Australia, 3 World Health Organization, 1211 Geneva 27, Switzerland, 4 School of Population Health, University of Queensland, Herston, QLD 4006, Australia
Correspondence to: M R Nelson Mark.Nelson@utas.edu.au
Objective To investigate the routine use of low dose aspirin in people aged 70 without overt cardiovascular disease.
Design Epidemiological modelling in a hypothetical population.
Setting Reference populations of men and women in the year 2000 from the state of Victoria, Australia.
Subjects 10 000 men and 10 000 women aged 70-74 with no cardiovascular disease.
Main outcome measures First ever myocardial infarction or unstable angina, ischaemic or haemorrhagic stroke, and major gastrointestinal haemorrhage. Health adjusted years of life lived.
Results The proportional benefit gained from the use of low dose aspirin by the prevention of myocardial infarctions (-389 in men, -321 in women) and ischaemic stroke (-19 in men and -35 in women) is offset by excess gastrointestinal (499 in men, 572 in women) and intracranial (76 in men, 54 in women) bleeding. The results in health adjusted years of life lived (which take into account length and quality of life) are equivocal for aspirin causing net harm or net benefit.
Conclusion Epidemiological modelling suggests that any benefits of low dose aspirin on risk of cardiovascular disease in people aged 70 are offset by adverse events. These findings are tempered by wide confidence intervals, indicating that the overall outcome could be beneficial or adverse.
The effects of low dose aspirin for the primary prevention of cardiovascular disease have been investigated in six large scale randomised clinical trials, which have been subjected to metaanalyses.1-3 On the basis of all but the last of these trials, current US guidelines recommend the use of low dose aspirin (75-150 mg) for people with a five year absolute coronary risk of 3% or a 10 year absolute cardiovascular risk of 10%.2 4 If implemented, these recommendations would mean that most elderly people would be prescribed aspirin prophylaxis because age is the greatest determinant of absolute risk. From the Australian diabetes, obesity, and lifestyle (AusDiab) data we estimated that in Australia about two thirds of people aged 70-74 (94% of men and 46% of women) have an estimated 10 year absolute cardiovascular risk of 10%.5 6 Prophylactic use of a potentially toxic agent can be problematic, however, particularly in people in whom comorbidity and polypharmacy are common.
In a prospective observational study in two large UK general hospitals, aspirin was the causal agent in 18% of all admissions for adverse drug reactions and was implicated in 61% of all associated deaths.7 Importantly, patients admitted with adverse drug reactions were significantly more likely to be older and female than those admitted without adverse drug reactions. In contrast, the primary prevention clinical trials were conducted mostly in middle aged people. The potential health gains of any preventive strategy need to be carefully balanced against their potential risks.
We simulated the broad implications of routine use of aspirin in patients aged 70. We used epidemiological modelling, a relatively new research method using evidence from clinical trials, descriptive epidemiology, and demography to critically inform clinical and public health practice.8
Methods
Model
We developed an epidemiological model in the configuration of a decision analysis tree, with the main branches representing the treatment options being compared.9 The progression of hypothetical cohorts of individuals through the decision analysis tree was underpinned by Markov modelling.10 This enables repeated analyses of the tree, which in essence allows for simulation of a period of follow-up during which multiple events can occur and the risks of these events change with time.
In our model we followed up 10 000 men and 10 000 women from the ages of 70-74 until death or 100. The figure shows the generic framework of the models. The tree is fully displayed only for the baseline arm (no aspirin treatment) but the treatment arm (aspirin treatment) has the same detail. In the baseline year (2000) everyone in the modelled cohort started in the state "alive before incident acute coronary syndrome and stroke." From there, probabilities of fatal and non-fatal disease, specific for age and sex, determined who made transitions to other health states (for example, alive after stroke or death) over time. We assumed that people who suffered episodes of non-fatal major gastrointestinal haemorrhage made full recovery.
Basic structure of model for cardiovascular disease prediction and estimation of the effects of low dose aspirin
To undertake comparative analyses of the two treatment strategies, we applied relative changes to the risks of incident acute coronary syndrome, haemorrhagic stroke, and major gastrointestinal bleeding, as indicated by the meta-analysis by Hayden et al,2 to the relevant transition probabilities. We derived the relative change to the risk of ischaemic stroke from the meta-analysis by Hart et al.1
The consequences of discontinuing treatment (for instance, people who did not comply with treatment or who were taken off aspirin after a major bleeding event) were modelled by stratification of the strategy arm into "on treatment" and "off treatment" branches (see figure). The model was also run assuming complete compliance with treatment therapy.
Outcomes of interest were lifetime differences between the treatment groups in terms of:
Fatal and non-fatal myocardial infarction/unstable angina
Fatal and non-fatal ischaemic stroke
Fatal and non-fatal haemorrhagic stroke
Fatal and non-fatal major gastrointestinal haemorrhage
Total years of life lived
Years of life lived adjusted for health.
We calculated health adjusted years of life lived by adjusting the years of life lived by a "disability weight" to reflect the disability associated with (non-fatal) health states.11 Future health gains were discounted to reflect society's preference for immediate rather than future health. The discount rate applied was 3% as recommended by the US Panel on Cost-effectiveness in Health and Medicine.12
To reflect uncertainty surrounding data inputs (and hence outputs) for the model, we entered these as ranges rather than single values. Each range was described by a probability distribution to reflect the nature of uncertainty. We used Monte Carlo simulation, with 2000 simulations for each analysis (that is, for each analysis, the progress of a cohort of 10 000 individuals was simulated 2000 times).13 During each simulation, we sampled a value from every input range according to its probability distribution. For each outcome of interest, the model generated 2000 results, and we derived uncertainty ranges from the distributions of these. Sensitivity analyses were also undertaken to determine which of the input ranges most influenced the modelled outputs.
The models were developed in Microsoft Excel with the software macro @Risk (Palisade Corporation, NY, USA) for Monte Carlo micro-simulation.
Data sources
The reference populations were men and women aged 70-74 in the year 2000 from Victoria, Australia. We used baseline population data from the Australian Bureau of Statistics and data on the prevalence of coronary heart disease and stroke, specific for age and sex, from the AusDiab study.5 14 The AusDiab study (1999-2001) collected representative cross sectional data on major chronic diseases and their risk factors in 11 427 Australians aged 25.
Incident rates of the outcomes of interest were derived from combined data from the Victorian admitted episodes database (VAED), the World Health Organization's monitoring trends and determinants of cardiovascular disease (MONICA) studies (two Australian sites), and the north east Melbourne stroke incidence study (NEMESIS).15-17 The VAED provides a comprehensive record of demographic and clinical information on all admissions to all public and private healthcare institutions across Victoria, and the NEMESIS study maintained a register of strokes occurring in a defined area of Melbourne (the capital city of Victoria) from the mid to the late 1990s.
Admissions for acute coronary syndrome were defined by ICD-10 (international classification of diseases, 10th edition) codes I20 (angina pectoris), I21 (acute myocardial infarction), I23 (certain current complications after acute myocardial infarction), and I24 (other acute ischaemic heart diseases). Admissions for stroke were defined by ICD-10 codes I61 (intracerebral haemorrhage), I62 (other non-traumatic intracranial haemorrhage), I63 (cerebral infarction), and I66 (stroke not specified as haemorrhagic or infarction). These codes had to have been nominated as one of the primary causes of admission.
We applied decreasing trends to transition probabilities underlying acute coronary syndrome and ischaemic and haemorrhagic stroke to reflect probable ongoing falls in the incidence and case fatality. These trends were extrapolated from those observed in Victoria from 1979 to 2000.18 Between 1981 and 2000 in the UK, mortality from coronary heart disease fell by 62% in men and 45% in women; 42% of the fall was attributed to individual treatment (including 11% to secondary prevention, 13% to heart failure treatment, 8% due to initial treatment of acute myocardial infarction, and 3% to blood pressure treatments) and 58% to reductions in population risk factors.19 On the basis of this finding, we apportioned 58% and 42% of the overall decreasing mortality trend to incidence and case fatality, respectively, in the model.
The use of evidence from trials for estimating gastrointestinal bleeding was problematic because most trials studied younger age groups, and results probably underestimate the risk in elderly people. To estimate rates of "aspirin modifiable" gastrointestinal bleeding, we first determined the rates of hospital admissions for gastrointestinal bleeding where there was no mention of cancer, cirrhosis/portal hypertension, vascular malformations in the gastrointestinal tract, or inflammatory bowel disease as underlying causes. These were then reduced proportionally by the number of people with existing coronary heart disease or stroke on low dose aspirin (using age and sex specific data on the prevalence of coronary heart disease and stroke reported in the AusDiab study) and by the greater share of risk for gastrointestinal bleeding (relative risk as indicated by the meta-analysis by Hayden et al) that they would have contributed.2
We used disability weights for coronary heart disease, stroke, and major gastrointestinal haemorrhage—ranging from 0 (perfect health) to 1 (worst possible health)—from the Australian burden of disease study.11 Projected years lived were further adjusted for other causes of ill health, with estimated prevalence and age specific weighted probabilities of disability from all other diseases apart from coronary heart disease and stroke. The contemporary publication of the women's health study (WHS) allowed us to recalculate outcomes for women using these data.
Inputs for the analysis
Tables 1, 2, 3 list all data inputs in the analyses.
Table 1 Baseline incidence rates (per person per year) of first ever events
Table 2 Baseline case fatality of first ever events. Figures represent triangular distributions specified as minimum, most likely value, maximum
Table 3 Other data inputs for modelling of effects of aspirin
Results
Table 4 summarises the results from the simulations. The model suggests that the benefit gained from routinely prescribing low dose aspirin to patients aged 70 in terms of preventing first ever coronary heart disease events would be offset by a greater occurrence of gastrointestinal and intracerebral bleeding. On balance, there was no indication of a net benefit or harm in terms of deaths, years of life saved, or years of healthy life saved. The last measure takes into account both length and quality of life and is therefore the most comprehensive measure of health effect.
Table 4 Simulated lifetime effects of low dose aspirin compared with no aspirin on cohorts of 10 000 men and 10 000 women in Australia, initially aged 70-74 years and free from cardiovascular disease. Figures are point estimates with 95% uncertainty intervals
Sensitivity analyses indicated dominance of the relative risks for disease associated with aspirin on the modelled outputs. For each of the disease related outcomes, whether incident events or deaths, the input variable that singularly dominated the modelled output was the relative risk of that particular outcome associated with aspirin. For example, the relative risk associated with coronary heart disease was the input variable which most influenced the predicted number of coronary heart disease events or deaths prevented. For the combined outcomes (years of life saved and health adjusted years of life saved), the relative risks associated with each of coronary heart disease, ischaemic stroke, and haemorrhagic stroke were the three most influential variables on the modelled outputs (to roughly equal extents), but the relative risk associated with major gastrointestinal haemorrhage was less influential.
When we assumed 100% compliance and used data from the women's health study, the outcomes were altered but not the overall uncertainty of the balance of events prevented versus adverse events (numbers not shown).
Discussion
Hart RG, Halperin JL, McBride R, Benavente O, Man-Son-Hing M, Kronmal RA. Aspirin for the primary prevention of stroke and other major vascular events: meta-analysis and hypotheses. Arch Neurol 2000;57: 326-32.
Hayden M, Pignone M, Phillips C, Mulrow C. Aspirin for the primary prevention of cardiovascular events: a summary of the evidence for the US preventive services task force. Ann Intern Med 2002;136: 161-72.
Ridker PM, Cook NR, Lee I-M, Gordon D, Gaziano JM, Manson JE, et al. A randomized trial of low-dose aspirin in the primary prevention of cardiovascular disease in women. N Engl J Med 2005;352: 1296-304.
Pearson T, Blair S, Daniels S, Eckel RH, Fair JM, Fortmann SP, et al. Guidelines for primary prevention of cardiovascular disease and stroke: 2002 update. Circulation 2002;106: 388-91.
Dunstan DW, Zimmet PZ, Welborn TA, Cameron AJ, Shaw J, de Courten M, et al. The Australian diabetes, obesity and lifestyle study (AusDiab)—methods and response rates. Diabetes Res Clin Pract 2002;57: 119-29.
D'Agostino RB, Russell MW, Huse DM, Ellison RC, Silbershatz H, Wilson PW, et al. Primary and subsequent coronary risk appraisal: new results from the Framingham study. Am Heart J 2000;139: 272-81.
Pirmohamed M, James S, Meakin S, Green C, Scott AK, Walley TJ, et al. Adverse drug reactions as cause of admission to hospital: prospective analysis of 18 820 patients. BMJ 2004;329: 15-9.
Liew D, McNeil JJ, Peeters A, Lim SS, Vos T. Epidemiological modelling (including economic modelling) and its role in preventive drug therapy. Med J Aust 2002;177: 364-7.
Lilford RJ, Pauker SG, Braunholtz DA, Chard J. Decision analysis and the implementation of research findings. BMJ 1998;317: 405-9.
Sonnenberg FA, Beck JR. Markov models in medical decision making: a practical guide. Med Decis Making 1993;13: 322-38.
Mathers C, Vos T, Stevenson C. The burden of disease and injury in Australia. Canberra: Australian Institute of Health and Welfare, 1999. (AIHW Cat No PHE 17.)
Weinstein MC, Siegel JE, Gold MR, Kamlet MS, Russell LB. Recommendations of the panel on cost-effectiveness in health and medicine. JAMA 1996;276: 1253-8.
Briggs AH. Handling uncertainty in cost-effectiveness models. Pharmacoeconomics 2000;17: 479-500.
Australian Bureau of Statistics. www.abs.gov.au/ (accessed April 2005).
Department of Human Services Victoria, Australia. Victorian admitted events dataset. 10th ed. Victoria, Australia: Department of Human Services, 2000/2001.
McElduff P, Dobson A, Jamrozik K, Hobbs M. The WHO MONICA study, Australia, 1984- 93: a summary of the Newcastle and Perth MONICA projects. Canberra: Australian Institute of Health and Welfare, 2000. (AIHW Cat No CVD 11.)
Thrift A, Dewey H, Macdonell R, McNeil JJ, Donnan GA. Stroke incidence on the east coast of Australia: the north east Melbourne stroke incidence study (NEMESIS). Stroke 2000;31: 2087-92.
Victorian Department of Human Services, Public Health, and Development Division. The Victorian burden of disease study: mortality. Victorian burden of disease study: morbidity. Melbourne: Victorian DHS, 1999.
Unal B, Critchley JA, Capewell S. Explaining the decline in coronary heart disease mortality in England and Wales between 1981 and 2000. Circulation 2004;109: 1101-7.
ASPREE Study Group, Nelson MR, Reid CM, Beilin LJ, Donnan GA, Johnston CI, et al. Rationale for a primary prevention trial of low dose aspirin for major adverse cardiovascular events and vascular dementia in the elderly. ASPirin in reducing events in the elderly (ASPREE). Drugs Aging 2003;20: 897-903.
Peto R, Gray R, Collins R, Wheatley K, Hennekens C, Jamrozik K, et al. Randomised trial of prophylactic daily aspirin in British male doctors. BMJ 1988;296: 313-6.
Steering Committee of the Physicians' Health Study Research Group. Final report on the aspirin component of the ongoing Physicians' Health Study. N Engl J Med 1989;321: 129-35.
Collaborative Group of the Primary Prevention Project (PPP). Low-dose aspirin and vitamin E in people at cardiovascular risk: a randomised trial in general practice. Lancet 2001;357: 89-95.
Hansson L, Zanchetti A, Carruthers S, Dahlof B, Elmfeldt D, Julius S, et al. Effects of intensive blood-pressure lowering and low dose aspirin in patients with hypertension: principal results of the hypertension optimal treatment (HOT) randomised trial. Lancet 1998;351: 1755-62.
Medical Research Council's General Practice Research Framework. Thrombosis prevention trial: randomised trial of low-intensity oral anticoagulation with warfarin and low-dose aspirin in the primary prevention of ischaemic heart disease in men at increased risk. Lancet 1998;351: 233-41.
Ross R. Atherosclerosis—an inflammatory disease. N Engl J Med 1999;143: 683-91.
Sturmer T, Glynn R, Field T, Taylor JO, Hennekens CH. Aspirin use and cognitive function in the elderly. Am J Epidemiol 1996;143: 683-91.
in t' Veld BA, Ruitenberg A, Hofman A, Launer LJ, van Duijn CM, Stijnen T, et al. Nonsteroidal antiinflammatory drugs and the risk of Alzheimer's disease. New Engl J Med 2001;345: 1515-21.
Baron JA, Sandler RS. Nonsteroidal anti-inflammatory drugs and cancer prevention. Annual Rev Med 2000;51: 511-23.
Antithrombosis Trialists' Collaboration. Collaborative meta-analysis of randomised trials of antiplatelet therapy for prevention of death, myocardial infarction, and stroke in high risk patients. BMJ 2002;324: 71-86.
Bonita R, Solomon N, Broad JB. Prevalence of stroke and stroke-related disability. Estimates from the Auckland stroke studies. Stroke 1997;28: 1898-902.(Mark R Nelson, chair1, Danny Liew, senio)
Correspondence to: M R Nelson Mark.Nelson@utas.edu.au
Objective To investigate the routine use of low dose aspirin in people aged 70 without overt cardiovascular disease.
Design Epidemiological modelling in a hypothetical population.
Setting Reference populations of men and women in the year 2000 from the state of Victoria, Australia.
Subjects 10 000 men and 10 000 women aged 70-74 with no cardiovascular disease.
Main outcome measures First ever myocardial infarction or unstable angina, ischaemic or haemorrhagic stroke, and major gastrointestinal haemorrhage. Health adjusted years of life lived.
Results The proportional benefit gained from the use of low dose aspirin by the prevention of myocardial infarctions (-389 in men, -321 in women) and ischaemic stroke (-19 in men and -35 in women) is offset by excess gastrointestinal (499 in men, 572 in women) and intracranial (76 in men, 54 in women) bleeding. The results in health adjusted years of life lived (which take into account length and quality of life) are equivocal for aspirin causing net harm or net benefit.
Conclusion Epidemiological modelling suggests that any benefits of low dose aspirin on risk of cardiovascular disease in people aged 70 are offset by adverse events. These findings are tempered by wide confidence intervals, indicating that the overall outcome could be beneficial or adverse.
The effects of low dose aspirin for the primary prevention of cardiovascular disease have been investigated in six large scale randomised clinical trials, which have been subjected to metaanalyses.1-3 On the basis of all but the last of these trials, current US guidelines recommend the use of low dose aspirin (75-150 mg) for people with a five year absolute coronary risk of 3% or a 10 year absolute cardiovascular risk of 10%.2 4 If implemented, these recommendations would mean that most elderly people would be prescribed aspirin prophylaxis because age is the greatest determinant of absolute risk. From the Australian diabetes, obesity, and lifestyle (AusDiab) data we estimated that in Australia about two thirds of people aged 70-74 (94% of men and 46% of women) have an estimated 10 year absolute cardiovascular risk of 10%.5 6 Prophylactic use of a potentially toxic agent can be problematic, however, particularly in people in whom comorbidity and polypharmacy are common.
In a prospective observational study in two large UK general hospitals, aspirin was the causal agent in 18% of all admissions for adverse drug reactions and was implicated in 61% of all associated deaths.7 Importantly, patients admitted with adverse drug reactions were significantly more likely to be older and female than those admitted without adverse drug reactions. In contrast, the primary prevention clinical trials were conducted mostly in middle aged people. The potential health gains of any preventive strategy need to be carefully balanced against their potential risks.
We simulated the broad implications of routine use of aspirin in patients aged 70. We used epidemiological modelling, a relatively new research method using evidence from clinical trials, descriptive epidemiology, and demography to critically inform clinical and public health practice.8
Methods
Model
We developed an epidemiological model in the configuration of a decision analysis tree, with the main branches representing the treatment options being compared.9 The progression of hypothetical cohorts of individuals through the decision analysis tree was underpinned by Markov modelling.10 This enables repeated analyses of the tree, which in essence allows for simulation of a period of follow-up during which multiple events can occur and the risks of these events change with time.
In our model we followed up 10 000 men and 10 000 women from the ages of 70-74 until death or 100. The figure shows the generic framework of the models. The tree is fully displayed only for the baseline arm (no aspirin treatment) but the treatment arm (aspirin treatment) has the same detail. In the baseline year (2000) everyone in the modelled cohort started in the state "alive before incident acute coronary syndrome and stroke." From there, probabilities of fatal and non-fatal disease, specific for age and sex, determined who made transitions to other health states (for example, alive after stroke or death) over time. We assumed that people who suffered episodes of non-fatal major gastrointestinal haemorrhage made full recovery.
Basic structure of model for cardiovascular disease prediction and estimation of the effects of low dose aspirin
To undertake comparative analyses of the two treatment strategies, we applied relative changes to the risks of incident acute coronary syndrome, haemorrhagic stroke, and major gastrointestinal bleeding, as indicated by the meta-analysis by Hayden et al,2 to the relevant transition probabilities. We derived the relative change to the risk of ischaemic stroke from the meta-analysis by Hart et al.1
The consequences of discontinuing treatment (for instance, people who did not comply with treatment or who were taken off aspirin after a major bleeding event) were modelled by stratification of the strategy arm into "on treatment" and "off treatment" branches (see figure). The model was also run assuming complete compliance with treatment therapy.
Outcomes of interest were lifetime differences between the treatment groups in terms of:
Fatal and non-fatal myocardial infarction/unstable angina
Fatal and non-fatal ischaemic stroke
Fatal and non-fatal haemorrhagic stroke
Fatal and non-fatal major gastrointestinal haemorrhage
Total years of life lived
Years of life lived adjusted for health.
We calculated health adjusted years of life lived by adjusting the years of life lived by a "disability weight" to reflect the disability associated with (non-fatal) health states.11 Future health gains were discounted to reflect society's preference for immediate rather than future health. The discount rate applied was 3% as recommended by the US Panel on Cost-effectiveness in Health and Medicine.12
To reflect uncertainty surrounding data inputs (and hence outputs) for the model, we entered these as ranges rather than single values. Each range was described by a probability distribution to reflect the nature of uncertainty. We used Monte Carlo simulation, with 2000 simulations for each analysis (that is, for each analysis, the progress of a cohort of 10 000 individuals was simulated 2000 times).13 During each simulation, we sampled a value from every input range according to its probability distribution. For each outcome of interest, the model generated 2000 results, and we derived uncertainty ranges from the distributions of these. Sensitivity analyses were also undertaken to determine which of the input ranges most influenced the modelled outputs.
The models were developed in Microsoft Excel with the software macro @Risk (Palisade Corporation, NY, USA) for Monte Carlo micro-simulation.
Data sources
The reference populations were men and women aged 70-74 in the year 2000 from Victoria, Australia. We used baseline population data from the Australian Bureau of Statistics and data on the prevalence of coronary heart disease and stroke, specific for age and sex, from the AusDiab study.5 14 The AusDiab study (1999-2001) collected representative cross sectional data on major chronic diseases and their risk factors in 11 427 Australians aged 25.
Incident rates of the outcomes of interest were derived from combined data from the Victorian admitted episodes database (VAED), the World Health Organization's monitoring trends and determinants of cardiovascular disease (MONICA) studies (two Australian sites), and the north east Melbourne stroke incidence study (NEMESIS).15-17 The VAED provides a comprehensive record of demographic and clinical information on all admissions to all public and private healthcare institutions across Victoria, and the NEMESIS study maintained a register of strokes occurring in a defined area of Melbourne (the capital city of Victoria) from the mid to the late 1990s.
Admissions for acute coronary syndrome were defined by ICD-10 (international classification of diseases, 10th edition) codes I20 (angina pectoris), I21 (acute myocardial infarction), I23 (certain current complications after acute myocardial infarction), and I24 (other acute ischaemic heart diseases). Admissions for stroke were defined by ICD-10 codes I61 (intracerebral haemorrhage), I62 (other non-traumatic intracranial haemorrhage), I63 (cerebral infarction), and I66 (stroke not specified as haemorrhagic or infarction). These codes had to have been nominated as one of the primary causes of admission.
We applied decreasing trends to transition probabilities underlying acute coronary syndrome and ischaemic and haemorrhagic stroke to reflect probable ongoing falls in the incidence and case fatality. These trends were extrapolated from those observed in Victoria from 1979 to 2000.18 Between 1981 and 2000 in the UK, mortality from coronary heart disease fell by 62% in men and 45% in women; 42% of the fall was attributed to individual treatment (including 11% to secondary prevention, 13% to heart failure treatment, 8% due to initial treatment of acute myocardial infarction, and 3% to blood pressure treatments) and 58% to reductions in population risk factors.19 On the basis of this finding, we apportioned 58% and 42% of the overall decreasing mortality trend to incidence and case fatality, respectively, in the model.
The use of evidence from trials for estimating gastrointestinal bleeding was problematic because most trials studied younger age groups, and results probably underestimate the risk in elderly people. To estimate rates of "aspirin modifiable" gastrointestinal bleeding, we first determined the rates of hospital admissions for gastrointestinal bleeding where there was no mention of cancer, cirrhosis/portal hypertension, vascular malformations in the gastrointestinal tract, or inflammatory bowel disease as underlying causes. These were then reduced proportionally by the number of people with existing coronary heart disease or stroke on low dose aspirin (using age and sex specific data on the prevalence of coronary heart disease and stroke reported in the AusDiab study) and by the greater share of risk for gastrointestinal bleeding (relative risk as indicated by the meta-analysis by Hayden et al) that they would have contributed.2
We used disability weights for coronary heart disease, stroke, and major gastrointestinal haemorrhage—ranging from 0 (perfect health) to 1 (worst possible health)—from the Australian burden of disease study.11 Projected years lived were further adjusted for other causes of ill health, with estimated prevalence and age specific weighted probabilities of disability from all other diseases apart from coronary heart disease and stroke. The contemporary publication of the women's health study (WHS) allowed us to recalculate outcomes for women using these data.
Inputs for the analysis
Tables 1, 2, 3 list all data inputs in the analyses.
Table 1 Baseline incidence rates (per person per year) of first ever events
Table 2 Baseline case fatality of first ever events. Figures represent triangular distributions specified as minimum, most likely value, maximum
Table 3 Other data inputs for modelling of effects of aspirin
Results
Table 4 summarises the results from the simulations. The model suggests that the benefit gained from routinely prescribing low dose aspirin to patients aged 70 in terms of preventing first ever coronary heart disease events would be offset by a greater occurrence of gastrointestinal and intracerebral bleeding. On balance, there was no indication of a net benefit or harm in terms of deaths, years of life saved, or years of healthy life saved. The last measure takes into account both length and quality of life and is therefore the most comprehensive measure of health effect.
Table 4 Simulated lifetime effects of low dose aspirin compared with no aspirin on cohorts of 10 000 men and 10 000 women in Australia, initially aged 70-74 years and free from cardiovascular disease. Figures are point estimates with 95% uncertainty intervals
Sensitivity analyses indicated dominance of the relative risks for disease associated with aspirin on the modelled outputs. For each of the disease related outcomes, whether incident events or deaths, the input variable that singularly dominated the modelled output was the relative risk of that particular outcome associated with aspirin. For example, the relative risk associated with coronary heart disease was the input variable which most influenced the predicted number of coronary heart disease events or deaths prevented. For the combined outcomes (years of life saved and health adjusted years of life saved), the relative risks associated with each of coronary heart disease, ischaemic stroke, and haemorrhagic stroke were the three most influential variables on the modelled outputs (to roughly equal extents), but the relative risk associated with major gastrointestinal haemorrhage was less influential.
When we assumed 100% compliance and used data from the women's health study, the outcomes were altered but not the overall uncertainty of the balance of events prevented versus adverse events (numbers not shown).
Discussion
Hart RG, Halperin JL, McBride R, Benavente O, Man-Son-Hing M, Kronmal RA. Aspirin for the primary prevention of stroke and other major vascular events: meta-analysis and hypotheses. Arch Neurol 2000;57: 326-32.
Hayden M, Pignone M, Phillips C, Mulrow C. Aspirin for the primary prevention of cardiovascular events: a summary of the evidence for the US preventive services task force. Ann Intern Med 2002;136: 161-72.
Ridker PM, Cook NR, Lee I-M, Gordon D, Gaziano JM, Manson JE, et al. A randomized trial of low-dose aspirin in the primary prevention of cardiovascular disease in women. N Engl J Med 2005;352: 1296-304.
Pearson T, Blair S, Daniels S, Eckel RH, Fair JM, Fortmann SP, et al. Guidelines for primary prevention of cardiovascular disease and stroke: 2002 update. Circulation 2002;106: 388-91.
Dunstan DW, Zimmet PZ, Welborn TA, Cameron AJ, Shaw J, de Courten M, et al. The Australian diabetes, obesity and lifestyle study (AusDiab)—methods and response rates. Diabetes Res Clin Pract 2002;57: 119-29.
D'Agostino RB, Russell MW, Huse DM, Ellison RC, Silbershatz H, Wilson PW, et al. Primary and subsequent coronary risk appraisal: new results from the Framingham study. Am Heart J 2000;139: 272-81.
Pirmohamed M, James S, Meakin S, Green C, Scott AK, Walley TJ, et al. Adverse drug reactions as cause of admission to hospital: prospective analysis of 18 820 patients. BMJ 2004;329: 15-9.
Liew D, McNeil JJ, Peeters A, Lim SS, Vos T. Epidemiological modelling (including economic modelling) and its role in preventive drug therapy. Med J Aust 2002;177: 364-7.
Lilford RJ, Pauker SG, Braunholtz DA, Chard J. Decision analysis and the implementation of research findings. BMJ 1998;317: 405-9.
Sonnenberg FA, Beck JR. Markov models in medical decision making: a practical guide. Med Decis Making 1993;13: 322-38.
Mathers C, Vos T, Stevenson C. The burden of disease and injury in Australia. Canberra: Australian Institute of Health and Welfare, 1999. (AIHW Cat No PHE 17.)
Weinstein MC, Siegel JE, Gold MR, Kamlet MS, Russell LB. Recommendations of the panel on cost-effectiveness in health and medicine. JAMA 1996;276: 1253-8.
Briggs AH. Handling uncertainty in cost-effectiveness models. Pharmacoeconomics 2000;17: 479-500.
Australian Bureau of Statistics. www.abs.gov.au/ (accessed April 2005).
Department of Human Services Victoria, Australia. Victorian admitted events dataset. 10th ed. Victoria, Australia: Department of Human Services, 2000/2001.
McElduff P, Dobson A, Jamrozik K, Hobbs M. The WHO MONICA study, Australia, 1984- 93: a summary of the Newcastle and Perth MONICA projects. Canberra: Australian Institute of Health and Welfare, 2000. (AIHW Cat No CVD 11.)
Thrift A, Dewey H, Macdonell R, McNeil JJ, Donnan GA. Stroke incidence on the east coast of Australia: the north east Melbourne stroke incidence study (NEMESIS). Stroke 2000;31: 2087-92.
Victorian Department of Human Services, Public Health, and Development Division. The Victorian burden of disease study: mortality. Victorian burden of disease study: morbidity. Melbourne: Victorian DHS, 1999.
Unal B, Critchley JA, Capewell S. Explaining the decline in coronary heart disease mortality in England and Wales between 1981 and 2000. Circulation 2004;109: 1101-7.
ASPREE Study Group, Nelson MR, Reid CM, Beilin LJ, Donnan GA, Johnston CI, et al. Rationale for a primary prevention trial of low dose aspirin for major adverse cardiovascular events and vascular dementia in the elderly. ASPirin in reducing events in the elderly (ASPREE). Drugs Aging 2003;20: 897-903.
Peto R, Gray R, Collins R, Wheatley K, Hennekens C, Jamrozik K, et al. Randomised trial of prophylactic daily aspirin in British male doctors. BMJ 1988;296: 313-6.
Steering Committee of the Physicians' Health Study Research Group. Final report on the aspirin component of the ongoing Physicians' Health Study. N Engl J Med 1989;321: 129-35.
Collaborative Group of the Primary Prevention Project (PPP). Low-dose aspirin and vitamin E in people at cardiovascular risk: a randomised trial in general practice. Lancet 2001;357: 89-95.
Hansson L, Zanchetti A, Carruthers S, Dahlof B, Elmfeldt D, Julius S, et al. Effects of intensive blood-pressure lowering and low dose aspirin in patients with hypertension: principal results of the hypertension optimal treatment (HOT) randomised trial. Lancet 1998;351: 1755-62.
Medical Research Council's General Practice Research Framework. Thrombosis prevention trial: randomised trial of low-intensity oral anticoagulation with warfarin and low-dose aspirin in the primary prevention of ischaemic heart disease in men at increased risk. Lancet 1998;351: 233-41.
Ross R. Atherosclerosis—an inflammatory disease. N Engl J Med 1999;143: 683-91.
Sturmer T, Glynn R, Field T, Taylor JO, Hennekens CH. Aspirin use and cognitive function in the elderly. Am J Epidemiol 1996;143: 683-91.
in t' Veld BA, Ruitenberg A, Hofman A, Launer LJ, van Duijn CM, Stijnen T, et al. Nonsteroidal antiinflammatory drugs and the risk of Alzheimer's disease. New Engl J Med 2001;345: 1515-21.
Baron JA, Sandler RS. Nonsteroidal anti-inflammatory drugs and cancer prevention. Annual Rev Med 2000;51: 511-23.
Antithrombosis Trialists' Collaboration. Collaborative meta-analysis of randomised trials of antiplatelet therapy for prevention of death, myocardial infarction, and stroke in high risk patients. BMJ 2002;324: 71-86.
Bonita R, Solomon N, Broad JB. Prevalence of stroke and stroke-related disability. Estimates from the Auckland stroke studies. Stroke 1997;28: 1898-902.(Mark R Nelson, chair1, Danny Liew, senio)