Appendix: Details of Calculation of Attributable Declines in Mortality

Estimating the contribution of the medical advances described in this book to the decline in heart attack mortality has been one of my recurring themes in this book. Ford et al. devised a model called IMPACT for this purpose, but the published results are now 20 years old and out of date.¹ Given the continuing advances in treatment and the continuing decline in heart attack mortality since 2000, it was imperative to extrapolate these results forward to a more current date, within the limitations of publicly available national health statistics. The year for which the necessary data were available ranged from 2009 for the in-hospital mortality data for acute myocardial infarction to 2017 for smoking and secondary prevention. I will provide the details of how these calculations were done in this appendix for those with a working knowledge of algebra, including natural logarithms (ln) and exponentials (exp).

I have made the following assumptions for these calculations:

1. I can identify a reported health statistic that can serve as a credible surrogate for the national state of treatment/control of each risk factor. For example, the percent of hypertensive patients whose BP is treated to < 140/90 mmHg for blood pressure treatment, the proportion of people on statins for cholesterol lowering, etc.

2. I have assumed a log-linear relationship between the surrogate metric and CHD mortality. In other words,

3. ln (CHD mortality ratio) for a given year versus baseline = β times the change in the surrogate metric

Table A.1 shows the national CHD (heart attack) mortality rates for the years that I used in my calculations.² The percent decline in mortality from 1968 and 1980 (the starting point for the published IMPACT model) are also shown.

Table A.1: Selected U.S. CHD Mortality Rates

Approximately 41% ((345.2–186.8)/(482.6–92.9)) of the total decline in U.S. CHD mortality since 1968 took place in the period (1980–2000) covered by the published IMPACT analysis.

Primary Prevention

Table A.2 shows the how the national health statistics for hypertension control, statin use, and cigarette smoking were used to extrapolate the published IMPACT data to five decades of decline in cardiovascular mortality. The top line shows the proportion of the 46% decline in CHD mortality (from 345.2 deaths per 100,000 in 1980 to 186.8 deaths per 100,000 in 2000) attributed in the IMPACT publication to interventions for high LDL cholesterol, high BP, and cigarette smoking.³ The second line (which is obtained by multiplying the top line by 46%) is the actual decrease in CHD mortality attributable to improvements in each of these risk factors. We know from the literature that the improvement in statin use, BP control, and smoking since 1968 exceeded that for the 1980–2000 period covered by the published IMPACT results by a factor ( “change ratio”) of 2.05 for statins, 2.46 for BP, and 2.57 for smoking.⁴ The log-linear relationship of CHD mortality to each of our surrogate risk factors allows us to multiply the natural log of the CHD risk ratio for 2000 versus 1980 by the change ratio, take the exponential to convert the result back to a mortality ratio, and subtract from 100% to obtain the percent mortality reduction. For smoking, as an example, one calculates

100% − exp (2.57*ln (1 − .45)) = 13.2%

to obtain the percent of CHD deaths prevented by smoking in 2017 relative to the number who would have died if smoking had remained unchanged since 1968. Dividing this number by the observed 81% decline in CHD mortality, one calculates that 16.4% of this decline was due to smoking. Similarly, 27.7% of the total decline through 2011 was attributable to LDL cholesterol lowering and 26.7% to BP control.

Table A.2. Contributions of Primary Prevention to the Decline in CHD Mortality, 1968 to present

*The IMPACT model estimate of the percent of decline attributed to the risk factor times the 45.9% decline in CHD mortality in 1980–2000.

^The change in the risk factor surrogate from 1968 to the present divided by the change from 1980 to 2000.

†The exponential of the product of the change ratio times the natural log of the attributable percent of deaths saved in 1980–2000.

^◊The ratio of the attributable percent of CHD deaths saved in 1968–present to the total percent decline in CHD mortality during that period.

^~Removes the overlap in deaths prevented by treating each of the three risk factors

The last line of Table A.2 shows the combined contribution of these three primary prevention risk factors to the decline in CHD mortality. One cannot simply add the three contributions (which total 71%) without subtracting the overlap. For example, if the 27.7% impact of LDL cholesterol lowering and the 26.7% impact of BP lowering are independent, there is a 7.4% overlap (27.7% times 26.7%) of the CHD deaths prevented. The easiest formula for calculating the combined impact of the three primary preventions in Table A.2 without the overlaps is:

100%–(100%-27.7%)*(100%-26.7%)*(100%-16.4%) = 55.7%

This means that primary prevention has reduced CHD mortality by 45% (55.7% of 80.7%) since 1968.

Since we have reliable cigarette smoking data going back to 1900, we can use the same logic to estimate its contribution to the rise of cardiovascular mortality in 1900–68.⁵ For this purpose, we have to shift from CHD to all heart disease (HD) mortality, since pre–1950 national mortality statistics do not distinguish between CHD, heart failure, and congenital, rheumatic, hypertensive, and other forms of heart disease.⁶ We must also correct the IMPACT estimate of the relative risk associated with not smoking versus smoking from 0.72 to 0.50, since the former figure applies to smoking cessation, rather than the initiation of smoking.⁷ Table A.3 shows the result.

Table A.3: Contributions of Cigarettes to the Rise in Heart Disease (HD) Mortality, 1900 to 1968

*The ratio of changes in smoking during 1900–1968 versus 1980–2000.

^Assumes that HD mortality in non-smokers is 50% (not 72%) of that in smokers.

†The IMPACT model estimate of the percent of decline attributed to smoking times the 37.5% decline in HD mortality in 1980–2000.

^◊The exponential of the product of the change ratio times the natural log of the attributable percent of deaths saved in 1980–2000.

This calculation estimates that smoking may have been responsible for 27.7% of the doubling of HD mortality between 1900 and 1968.

Secondary Prevention

Table A.4 shows the calculation of the contributions of the four major secondary prevention drugs to the 81% decline in CHD mortality between 1968 and 2017. I have assumed that none of these drugs were in significant use in heart attack survivors before 1980; indeed statins and ACE inhibitors had not yet been invented. I have taken the 2000 drug usage data from the Ford IMPACT paper for secondary prevention after MI; the 2017 drug usage data are projections from a 2012 analysis of NHANES data by Shah, et al.⁸ The logic of the table follows that of Table A.2, and I will not repeat it here.

Table A.4. Contributions of Secondary Prevention to the Decline in CHD Mortality, 1968 to Present

*The IMPACT model estimate of the percent of decline attributed to the risk factor times the 45.9% decline in CHD mortality in 1980–2000.

^The change in the risk factor surrogate from 1968 to the present divided by the change from 1980 to 2000.

†The exponential of the product of the change ratio times the natural log of the attributable percent of deaths saved in 1980–2000.

^◊The ratio of the attributable percent of CHD deaths saved in 1968–present to the total percent decline in CHD mortality during that period.

^~Removes the overlap in deaths prevented by each of the four treatments.

The modified IMPACT estimates are quite similar to the published ones for statins and aspirin, indicating that their usage has increased more or less in parallel to the continuing decline in CHD mortality since 2000. The usage of beta-blockers has not kept pace with the other drugs, while increases in the use of RAAS inhibitors have modestly outpaced those in statins and aspirin. These four secondary prevention drugs, considered together, account for 23.3% of the 81% decline in CHD mortality since 1968—i.e., a net 18.8% reduction in CHD mortality. As for primary prevention (Table A.2), this is less than the sum of their individual contributions (25.7%); the difference reflects the overlap of deaths prevented by each drug.

I do not have confidence in the IMPACT model’s estimates of the contribution of improvements in the treatment of acute myocardial infarction and diabetes, and the (negative) contribution of the rapid increase in the prevalence of type 2 diabetes, since both occurred largely after 2000. Instead, I have made my own estimates.

Type 2 Diabetes

I have assumed that type 2 diabetes still confers a threefold increase in risk of CHD mortality relative to non-diabetics and that the prevalence of type 2 diabetes was 8.2% in U.S. adults as of 2017.⁹ Note that the threefold relative risk reflects similar and substantial reductions in CHD mortality of diabetic and non-diabetic persons due to the use of statins, BP drugs, etc., that were unavailable in 1968. So let us use M to symbolize the CHD mortality in deaths per 100,000. p to symbolize the prevalence of diabetes, X to symbolize the CHD mortality rate in non-diabetic persons, and 3x to symbolize the CHD mortality rate in diabetics. We can write the following equation and solve for x:

M = (1-p)*x + 3px

x = M/(2p+1)

The results for 1968 and 2017 are tabulated in lines 2–3 of Table A.5.

Table A.5. CHD Mortality Projections for Diabetes

In the last three lines of the table, the CHD mortality rates for 1968 and 2017 are recalculated under three hypothetical scenarios. First, if the prevalence of diabetes had been 8.2% (instead of 1.6%) in 1968 and nothing else changed, the CHD mortality rate would have been 544.3 instead of 482.6 deaths per 100,000. Conversely, if the prevalence of diabetes had remained at 1.6% (instead of 8.2%) in 2017, the CHD mortality rate would have been 82.4 instead of 92.9 deaths per 100,000. Thus, if the prevalence of diabetes had not changed between 1968 and 2017, CHD mortality decreased by 82.9% (from 482.6 to 82.4 or from 544.3 to 92.9 deaths per 100,000), rather than by 80.7%.

Finally, in the last line of Table A.5, we imagine a worst case scenario, in which the prevalence of diabetes stays on its current trajectory and reaches 40% sometime around 2040, while nothing else changes. In that scenario, CHD mortality would increase by 55% of its 2017 rate to reach 143.7 deaths per 100,000 annually.

Acute MI

Table A.6 lays out the calculation of the contribution of PCI and other acute interventions to the 73.7% decline in CHD between 1970 and 2009, the timespan for which published in-hospital case-fatality ratios for acute MI are available.

Table A.6. Contribution of Improved In-Hospital Survival to the Decline in Heart Attack Mortality, 1970–2009

*Age-adjusted to the U.S. population in 2000.

^Assumes that 55.7% of the decline in heart attack deaths and acute MI hospitalizations in 1970 could have been prevented if the primary prevention measures in effect in 2009 had been available in 1970.

The first line of Table A.6 shows the national heart attack death rates for 1970 and 2009 from Table A.1, which decreased from 448.0 to 117.7—a decline of 330.3 deaths per 100,000. In the second line, I have adjusted the 1970 rate to remove the impact of primary prevention by subtracting 55.7% of this 330.3 deaths per 100,000 decline—the proportion attributable to primary prevention in Table A.2—from the unadjusted 448.0 figure, to obtain an adjusted rate of 264.1 deaths per 100,000 in 1970. This figure represents what the mortality rate would have been if the 2009 levels of smoking and BP and LDL cholesterol treatment had prevailed in 1970. The third line shows the national incidence rates of hospitalized acute MI, adjusted to the 2000 U.S. Population using census data for persons over age 45, which have decreased by 88.7 cases per 100,000 between 1970 and 2009.¹⁰ You will note that this decrease between 1970 and 2009 is far smaller than the decrease in CHD mortality rates; this reflects the use of more sensitive diagnostic criteria for acute MI—specifically high-sensitivity serum troponin levels—in 2009.¹¹ In line 4, I subtracted 55.7% of the 88.7 case difference in acute MI hospitalizations from 268.4 to obtain an adjusted rate of 219.0 cases per 100,000 for 1970. Now, applying the reported age-adjusted case-fatality rates for ages > 45 (line 5), which dropped from almost 23.9% in 1970 to only 4.0% in 2009, to the adjusted incidence rates (line 4), one calculates a decrease of 45.2 deaths per 100,000 (line 7) in the national in-hospital mortality rate after adjustment for primary prevention.¹² This represents 13.7% of the 330.3 deaths per thousand unadjusted difference in overall CHD mortality between 1970 and 2009.

1. ES Ford, UA Ajani, JB Croft, JA Critchley, DR Labarth, TE Kottke, WH Giles, S Capewell. Explaining the Decrease in U.S. Deaths from Coronary Disease, 1980–2000. N Engl J Med 2007; 356:2388–2398. DOI: 10.1056/NEJMsa 053935.

2. Centers for Disease Control and Prevention (CDC), National Center of Health Statistics. Mortality Data Finder. Table 5: Age-adjusted death rates for selected causes of death by sex, race and Hispanic origin: United States, selected years 1950–2017 https://www.cdc.gov/nchs/hus/contents2018.htm#Table_005 (Excel spreadsheet link).

Morbidity and Mortality. 2012 Chartbook on Cardiovascular, Lung and Blood Diseases, NIH-NHLBI. Chart 3–24. https://www.nhlbi.nih.gov/files/docs/research/2012_ChartBook_508.pdf.

Age-adjusted death rates for 69 selected causes by race and sex using year 2000 standard population: United States, 1968–78 https://www.cdc.gov/nchs/data/mortab/aadr6878.pdf.

3. ES Ford et al.

4. EM Sarpong, SH Zuvekas. Changes in statin therapy among adults (age 18+) by selected characteristics, United States, 2000–2001 to 2010–2011. Medical Expenditure Panel Survey (MEPS). Statistical Brief #459. November 2014. https://www.ncbi.nlm.nih.gov/books/NBK470833/.

VL Burt, JA Cutler, M Higgins, MJ Horan, D Labarth, P Whelton, C Brown, EJ Rocella. Trends in the prevalence, awareness, treatment, and control of hypertension in the adults U.S. population. Data from the Health Examination Surveys, 1960 to 1991. Hypertension 1995; 26:1–60, https://www.ahajournals.org/doi/epub/10.1161/01.HYP.26.1.60.

P Muntner, ST Hardy, LJ Fine, BC Jaeger, G Wozniak, EB Levitan, LD Colantonio. Trends in blood pressure control among U.S. adults with hypertension, 1999–2000 to 2017–2018. JAMA 2020; doi:10.1001/jama.2020.14545.

American Lung Association. Overall Tobacco Trends. https://www.lung.org/research/trends-in-lung-disease/tobacco-trends-brief/overall-tobacco-trends. Accessed September 2020.

5. American Lung Association. Overall Tobacco Trends.

6. Centers for Disease Control and Prevention (CDC), National Center of Health Statistics Mortality Data, HIST293. Age-adjusted death rates for selected causes by race and sex using year 2000 standard population: death registration states, 1900‐32 and United States, 1933‐49, Diseases of the Heart. https://www.cdc.gov/nchs/data/dvs/hist293_1900_49.pdf. Morbidity and Mortality. 2012.

Centers for Disease Control and Prevention (CDC), National Center of Health Statistics. Mortality Data Finder. Table 5: Age-adjusted death rates for selected causes of death by sex, race and Hispanic origin: United States, selected years 1950–2017 https://www.cdc.gov/nchs/hus/contents2018.htm#Table_005 (Excel spreadsheet link).

7. JA Critchley, S Capewell. Mortality risk reduction associated with smoking cessation in patients with coronary heart disease: a systematic review. JAMA 2003; 290:86–97. Doi: 10.1001/jama.290.1.86.

8. ES Ford et al.

NS Shah, MD Huffman, H Ning, DM Lloyd-Jones. Trends in myocardial infarction secondary prevention: The National Health and Nutrition Examination Surveys (NHANES), 1999–2012. J Am Heart Assoc 2015; 4:1–12. doi:10.1161/JAHA.114.001709. https://www.ahajournals.org/doi/pdf/10.1161/JAHA.114.001709.

9. CDC National Diabetes Statistical Report 2020. Estimates of Diabetes and its Burden in the United States. https://www.cdc.gov/diabetes/pdfs/data/statistics/national-diabetes-statistics-report.pdf

AHA/ACC Heart Risk Calculator, http://www.cvriskcalculator.com/.

10. Morbidity and Mortality. 2012 Chartbook on Cardiovascular, Lung and Blood Diseases, NIH-NHLBI. Chart 3–22. https://www.nhlbi.nih.gov/files/docs/research/2012_ChartBook_508.pdf.

LM Howden, JA Meyer. 2010 Census Briefs: Age and Sex Composition 2010. U.S. Department of Commerce, Economics and Statistics Information, U.S. Census Bureau. May 2011. https://www.census.gov/prod/cen2010/briefs/c2010br-03.pdf.

11. PA Kavsak, AR MacRae, V Lustig, R Bhargava, R Vandersluis, GE Polomaki, ML Yerna, AS Jaffe. The impact of the ESC/ACC redefinition of myocardial infarction and new sensitive troponin assays on the frequency of acute myocardial infarction. Am Heart J 2005; 152:118–125.

12. Morbidity and Mortality. 2012 Chartbook on Cardiovascular, Lung and Blood Diseases, NIH-NHLBI. Chart 3–23. https://www.nhlbi.nih.gov/files/docs/research/2012_ChartBook_508.pdf.

LM Howden, JA Meyer. 2010 Census Briefs.