Published on

Published in: Health & Medicine, Technology
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide


  1. 1. The impact of new drug launches on longevity: evidence from longitudinal, disease-level data from 52 countries, 1982-2001 Frank R. Lichtenberg Columbia University and   National Bureau of Economic Research
  2. 2. United Nations Human Development Index <ul><li>(unweighted) average of three indexes: </li></ul><ul><li>an index of per capita GDP </li></ul><ul><li>a life expectancy index </li></ul><ul><li>an education index </li></ul>
  3. 3. U.S. economic growth, 20 th C. Nordhaus: “to a first approximation, the economic value of increases in longevity over the twentieth century is about as large as the value of measured growth in non-health goods and services”
  4. 4. Life expectancy at birth, world, 1950-2000
  5. 5. Life expectancy at birth, by region Unlike GDP, longevity is converging
  6. 6. Sources of longevity increase? <ul><li>improved quality of, and access to, medical care </li></ul><ul><li>other factors </li></ul>
  7. 7. Conventional wisdom <ul><li>“ the empirical evidence indicates [that] the overall contribution of medical care to health is rather modest at the margin …education, lifestyle, the environment, and income [are] the major contributing factors” (Santerre and Neun (2000, p. 69)). </li></ul><ul><li>“ increase in life expectancy [has] been much more influenced by economic development than improvements in medical care …the most important medical advances are being brought about by improvements in information technology, not pills and scalpels” (Getzen (1997, p. 330)). </li></ul>
  8. 8. Conventional wisdom <ul><li>“ Research on the relationship between health status and medical care frequently has found that the marginal contribution of medical care to health status is rather small …any significant improvements in health status are more likely to originate from factors other than medical care…Factors that determine the level of health include income and education, environmental and life-style factors, and genetics” (Henderson (1999, p.142)). </li></ul><ul><li>“ The historical declines in population mortality rates were not due to medical interventions because effective medical interventions became available to populations largely after the mortality had declined. Instead, public health, improved environment, and improved nutrition probably played substantial roles” (Folland, Goodman, and Stano (2001, p. 118)). </li></ul>
  9. 9. Paul Romer’s Model of Endogenous Technical Progress <ul><li>Y = (A L) 1-  K  </li></ul><ul><ul><li>Y = output </li></ul></ul><ul><ul><li>A = the “stock of ideas” </li></ul></ul><ul><ul><li>L = labor used to produce output </li></ul></ul><ul><ul><li>K = capital </li></ul></ul><ul><ul><li>0 <  < 1 </li></ul></ul><ul><li>The cumulative number of drugs launched (N_DRUG) is analogous to the stock of ideas. </li></ul>
  10. 10. Health production function <ul><li>AGE_DEATH ijt =  ln(N_DRUG ij,t-k ) </li></ul><ul><li> +  X ijt +  ijt </li></ul><ul><ul><li>AGE_DEATH ijt = a statistic based on the age distribution of deaths from disease i in country j in year t </li></ul></ul><ul><ul><li>N_DRUG ij,t-k = the number of drugs launched to treat disease i in country j by year t-k </li></ul></ul><ul><ul><li>X ijt = a vector of other factors (e.g. education, income, nutrition, the environment, and “lifestyle”) affecting the age distribution of deaths from disease i in country j in year t </li></ul></ul>
  11. 11. Specification <ul><li>diminishing returns to additions to the stock of drugs </li></ul><ul><li>specify a k-year lag in the relationship to allow for gradual diffusion of new drugs to consumers; we will estimate the model using different assumed values of k (k = 0, 1, 2,…). </li></ul>
  12. 12. Controlling for “other factors” <ul><li>Hypothesize that many of the “other factors” affecting the age distribution of deaths from disease i in country j in year t (e.g. per capita income, public health expenditure, and environmental quality) are: </li></ul><ul><li>invariant across diseases within a country and year </li></ul><ul><li>invariant across countries within a disease and year, or </li></ul><ul><li>invariant across years within a country and disease </li></ul>
  13. 13. Controlling for “other factors” <ul><li>decompose X ijt as follows: </li></ul><ul><li>  </li></ul><ul><li>X ijt =  ’ it +  ’ jt +  ’ ij +  ’ ijt (2) </li></ul><ul><li>  </li></ul><ul><li>where </li></ul><ul><li> ’ it = a fixed effect for disease i in year t </li></ul><ul><li> ’ jt = a fixed effect for country j in year t </li></ul><ul><li> ’ ij = a fixed effect for disease i in country j </li></ul>
  14. 14. Reduced form <ul><li>AGE_DEATH ijt =  ln(N_DRUG ij,t-k ) </li></ul><ul><li> +  it +  jt +  ij + u ijt </li></ul><ul><li>Zero-lag equation (k = 0), is estimated using 4678 observations, included 496 country*year effects, 189 disease*year effects, and 502 country*disease effects. The equations are estimated via weighted least squares, using the number of deaths in that disease-country-year cell as the weight. </li></ul>
  15. 15. IMS Health Drug Launches database <ul><li>Has tracked new product introductions worldwide since 1982 </li></ul><ul><li>In August 2001 the database contained over 165,000 records of individual product introductions between 1982 and 2001 </li></ul><ul><li>Allows measurement, for each country and therapeutic area, of the total number of ingredients launched, and the number of new chemical entities launched </li></ul>
  16. 16. Example: tenecteplase <ul><li>Launch date Country </li></ul><ul><li>6/00 USA </li></ul><ul><li>3/01 Finland </li></ul><ul><li>5/01 UK </li></ul><ul><li>9/01 Norway </li></ul><ul><li>10/01 Canada </li></ul><ul><li>10/01 South Africa </li></ul><ul><li>11/01 Ireland </li></ul>
  17. 17. Drug launch probability profiles: U.S. vs. Canada
  18. 18. Censoring of drug launches <ul><li>IMS Health Drug Launches database has tracked new product introductions worldwide since 1982 </li></ul><ul><li>NCE launches are guaranteed to be initial launches, but non-NCE launches may be either initial launches or re-launches; we suspect they are predominantly the latter. </li></ul>
  19. 19. Censoring of drug launches <ul><li>AGE_DEATH ijt =  NCE ln(CUM_NCE ij,t-k ) </li></ul><ul><li>+  NON ln(CUM_non-NCE ij,t-k ) </li></ul><ul><li>+  it +  jt +  ij + u ijt </li></ul><ul><li>CUM_NCE = the cumulative number of NCEs launched </li></ul><ul><li>CUM_non-NCE = the cumulative number of non-NCEs launched </li></ul><ul><li>Hypothesize that  NCE >  NON </li></ul><ul><li> NON could be negative? </li></ul>
  20. 20. WHO Mortality database <ul><li>Provides data on the age distribution of deaths, by disease, country, and year </li></ul><ul><li>Use aggregate life tables to translate our estimates of the impact of new drug launches on survival probabilities into estimates of the impact of new drug launches on life expectancy </li></ul>
  21. 21. Relationship between life expectancy and probability of survival to age 65, U.S., 1900-2000
  22. 22. Linkage of drug launches to diseases <ul><li>Drug launches documented in the IMS Health Drug Launches database are classified by therapeutic category </li></ul><ul><li>Deaths documented in the WHO Mortality Database are classified by cause (disease), using the International Classification of Diseases </li></ul><ul><li>The high-level IMS drug classification corresponds quite closely to the high-level ICD disease classification, e.g. cardiovascular system drugs obviously correspond to (are used to treat) diseases of the circulatory system </li></ul>
  23. 23. 11 broad disease categories
  24. 24. Countries with most and fewest drug launches
  25. 25. Findings <ul><li>Launches of New Chemical Entities (NCEs) have a strong positive impact on the probability of survival </li></ul><ul><li>It takes at least three years for new NCE launches to have their maximum impact on survival rates </li></ul><ul><li>This is probably due to the gradual diffusion of drugs to consumers following launch; data on pharmaceutical expenditure are consistent with this interpretation </li></ul><ul><li>Launches of (older) drugs that are not NCEs—many of which may already have been on the market—do not increase longevity </li></ul>
  26. 26. Estimates of  NCE for different lags between stock of NCEs launched and longevity
  27. 27. Estimates of  NCE and  expend at different lag values
  28. 28. Contribution of NCE launches to longevity increase <ul><li>NCE launches appear to account for a significant fraction of the long-run increase in longevity in the sample as a whole </li></ul><ul><li>Between 1986 and 2000, average life expectancy of the entire population of sample countries increased by almost two (1.96) years. </li></ul><ul><li>The estimates imply that NCE launches accounted for 0.79 years (40%) of the 1986-2000 increase in longevity. </li></ul><ul><li>The average annual increase in life expectancy of the entire population resulting from NCE launches is .056 years, or 2.93 weeks. </li></ul>
  29. 29. Contribution of NCE launches to increase in average life expectancy of the population since 1986
  30. 30. Cost per life-year gained from the launch of NCEs <ul><li>In 1997, average per capita pharmaceutical expenditure in OECD countries was about $250 </li></ul><ul><li>The average annual increase in life expectancy of the entire population resulting from NCE launches is .056 years </li></ul><ul><li>Hence pharmaceutical expenditure per person per year divided by the increase in life-years per person per year attributable to NCE launches is about $4500 </li></ul><ul><li>This is far lower than most estimates of the value of a life-year </li></ul><ul><li>Moreover, since the numerator includes expenditure on old drugs as well as on recently-launched NCEs, it probably grossly overstates the cost per life-year gained from the launch of NCEs </li></ul>
  31. 31. Micro evidence from a Medicaid program
  32. 32. Probability of death by end of 2002 <ul><li>Vintage of drugs </li></ul><ul><li>used Jan-June 2000 </li></ul><ul><li>% approved after 1970 </li></ul><ul><li>% approved after 1980 </li></ul><ul><li>% approved after 1990 </li></ul><ul><li>Other characteristics </li></ul><ul><li>age </li></ul><ul><li>sex </li></ul><ul><li>region </li></ul><ul><li>utilization Jan-June 2000 </li></ul><ul><ul><li>no. of MD visits </li></ul></ul><ul><ul><li>no. of Rx’s </li></ul></ul><ul><ul><li>no. of hospital admissions </li></ul></ul><ul><li>nature of person’s illnesses </li></ul>540,000 people 12.2 million claims
  33. 33. Data <ul><li>All medical and pharmacy claims of Medicaid beneficiaries during the period January 1-June 30, 2000 </li></ul><ul><ul><li>Almost 800,000 people; 540,000 had pharmacy claims </li></ul></ul><ul><ul><li>About 12.2 million claims </li></ul></ul><ul><li>List of all residents who died during the period 2000-2002. </li></ul>
  34. 34. Mortality rate declines as drug vintage increases
  35. 35. Actual vs. hypothetical mortality rates
  36. 36. Analysis by disease group
  37. 37. The Economics of Invention Incentives: Patents, Prizes, and Research Contracts Brian D. Wright American Economic Review 73, 1983, pp. 691-707.
  38. 38. How should the government support biomedical research? <ul><li>Alternative mechanisms: </li></ul><ul><li>Government labs </li></ul><ul><li>Research grants and contracts </li></ul><ul><li>Regulation (e.g. Orphan Drug Act) </li></ul><ul><li>Antitrust law (Joint ventures) </li></ul><ul><li>Patent law </li></ul><ul><li>Prizes & purchase commitments </li></ul>
  39. 39. Simple model of research <ul><li>Large number of firms, each of which can undertake one research project </li></ul><ul><li>Each research firm can conduct one research study at a cost of c = $1 </li></ul><ul><li>The more firms actively searching for a particular invention, the higher the probability that at least one of them will discover it. The probability of success is an increasing function of n </li></ul>
  40. 40. Optimal number of firms
  41. 41. Optimal number of firms
  42. 42. Research contracts <ul><li>If government can determine the optimal number of contracts (n), and firms engage in energetic research even though payments are independent of success, govt. should offer research contracts to the n lowest bidders; competition drives price down to cost </li></ul>
  43. 43. Government prizes <ul><li>Even if there is no patent protection, a large enough prize can induce research </li></ul><ul><li>If the government sets the prize properly, the optimal number of firms race to win it </li></ul><ul><li>A higher prize stimulates excessive research </li></ul><ul><li>A prize equal to the social value of the innovation may be too high—it induces excessive innovation </li></ul>
  44. 44. Government uncertainty <ul><li>When the government has full information, patents and joint ventures are less desirable than prizes or research contracts because they distort pricing </li></ul><ul><li>However, if inventors have more information before they start inventing than do government officials, patents and joint ventures may be superior </li></ul>
  45. 45. Patents <ul><li>Because patents lead to distortions due to monopoly pricing, they are less efficient than optimal prizes or research contracts if the government has sufficient information to induce the optimal amount of research </li></ul><ul><li>Permanent patent may lead to excessive research </li></ul><ul><li>By having patents last shorter periods of time, the government can reduce the incentive for excessive research </li></ul>
  46. 46. Patents <ul><li>Tradeoff: the longer the patent, the greater the inducement of research (and the probability of success), but the larger the cost due to more research projects and the monopoly loss </li></ul><ul><li>Government should choose patent length to maximize expected net social benefit </li></ul><ul><li>Because of the distortions associated with patents, society may want fewer projects than it would with prizes or research contracts </li></ul>
  47. 47. Public policy towards innovation <ul><li>In reality, the government cannot directly control the number of projects </li></ul><ul><li>But the government can influence the number of projects by establishing a system of intellectual property rights (e.g. patents), which affect firms’ incentives to invest in R&D </li></ul><ul><li>Designing an optimal patent system is a challenging task, however. A patent system could lead to either too little or too much R&D investment. </li></ul>
  48. 48. Patents: benefits and risks <ul><li>In the absence of patents, there may be inadequate investment in R&D, since firms attempt to “free ride” on other firms’ investments </li></ul><ul><li>Patents can solve the problem of under-investment. </li></ul><ul><li>However, since patents create a “winner-take-all” competition, patents can cause over-investment. </li></ul><ul><li>Other aspects of patents </li></ul><ul><ul><li>Prices </li></ul></ul><ul><ul><li>Disclosure of invention </li></ul></ul><ul><ul><li>Sequential innovation </li></ul></ul>