The impact of pharmaceutical innovation on premature cancer mortality in Canada, 2000–2011

The premature cancer mortality rate has been declining in Canada, but there has been considerable variation in the rate of decline across cancer sites. I analyze the effect that pharmaceutical innovation had on premature cancer mortality in Canada during the period 2000–2011, by investigating whether the cancer sites that experienced more pharmaceutical innovation had larger declines in the premature mortality rate, controlling for changes in the incidence rate. Premature mortality before age 75 is significantly inversely related to the cumulative number of drugs registered at least 10 years earlier. Since mean utilization of drugs that have been marketed for less than 10 years is only one-sixth as great as mean utilization of drugs that have been marketed for at least a decade, it is not surprising that premature mortality is strongly inversely related only to the cumulative number of drugs that had been registered at least ten years earlier. Premature mortality before age 65 and 55 is also strongly inversely related to the cumulative number of drugs that had been registered at least ten years earlier. None of the estimates of the effect of incidence on mortality are statistically significant. Controlling for the cumulative number of drugs, the cumulative number of chemical subgroups does not have a statistically significant effect on premature mortality. This suggests that drugs (chemical substances) within the same class (chemical subgroup) are not therapeutically equivalent. During the period 2000–2011, the premature (before age 75) cancer mortality rate declined by about 9 %. The estimates imply that, in the absence of pharmaceutical innovation during the period 1985–1996, the premature cancer mortality rate would have increased about 12 % during the period 2000–2011. A substantial decline in the “competing risk” of death from cardiovascular disease could account for this. The estimates imply that pharmaceutical innovation during the period 1985–1996 reduced the number of years of potential life lost to cancer before age 75 in 2011 by 105,366. The cost per life-year before age 75 gained from previous pharmaceutical innovation is estimated to have been 2730 USD. Most of the previously-registered drugs were off-patent by 2011, but evidence suggests that, even if these drugs had been sold at branded rather than generic prices, the cost per life-year gained would have been below 11,000 USD, a figure well below even the lowest estimates of the value of a life-year gained. The largest reductions in premature mortality occur at least a decade after drugs are registered, when their utilization increases significantly. This suggests that, if Canada is to obtain substantial additional reductions in premature cancer mortality in the future (a decade or more from now) at a modest cost, pharmaceutical innovation (registration of new drugs) is needed today. Electronic supplementary material The online version of this article (doi:10.1007/s10754-015-9172-2) contains supplementary material, which is available to authorized users.


Introduction
Previous authors have argued that "reducing premature mortality is a crucial public health objective" (Renard et al. 2014). A widely used measure of premature mortality is years of potential life lost (YPLL) before a given age (e.g. age 75), i.e. the number of years not lived by an individual who died before that age (Association of Public Health Epidemiologists in Ontario 2015). Statistics of YPLL are published by the World Health Organization, the OECD, and government agencies of Canada, the U.S., and other countries. Burnet et al. (2005) argue that YPLL "should be considered when allocating research funds." In the U.S., "cancer [was] responsible for more [YPLL] than all other causes of death combined" in 2008 (National Cancer Institute 2015c). In Canada, premature (before age 75) mortality from cancer is about twice as great as premature mortality from circulatory diseases.
But as shown in Fig. 1, the premature cancer mortality rate has been declining; it declined about 20 % between 1996 and 2006. The cancer incidence rate remained approximately constant during that period.
While the premature mortality rate from all cancers combined has been declining in Canada, Fig. 2 indicates that there has been considerable variation in the rate of decline across cancer sites. During the period 2000-2011, the premature mortality rate from breast cancer declined 20 %, and from cancer of lymphoid, haematopoietic and related tissue declined 27 %, but the premature mortality rate from lip, oral cavity, and pharynx cancer increased 6 %, and from cancer of female genital organs increased 8 %. I will show that this variation in the rate of decline of premature mortality cannot be explained by variation in the rate of decline of incidence.
In this paper, I will analyze the effect that pharmaceutical innovation has had on premature cancer mortality in Canada during the period 2000-2011. 1,2 The analysis will be performed using a difference-in-differences research design based on longitudinal disease-level data. 1 Lichtenberg (2014b) analyzed the impact of pharmaceutical innovation and other types of medical innovation on cancer mortality in the U.S. during the period [2000][2001][2002][2003][2004][2005][2006][2007][2008][2009]. But as the Squires (2011) demonstrated, the U.S. health care system differs dramatically from the health care systems of other OECD countries, including Canada. For example, in 2008 per capita spending on health was 85 % higher in the U.S. than it was in Canada. Also, the outcome measure and the measure of pharmaceutical innovation used in the present study will differ from those used in Lichtenberg (2014b The premature mortality rate is the number of potential years of life lost before age 75 per 100,000 population age 0-74. In essence, I will investigate whether the cancer sites that experienced more pharmaceutical innovation had larger declines in the premature mortality rate, controlling for changes in the incidence rate. Figure 3 illustrates that the rate of pharmaceutical innovation, as measured by the number of drugs registered during the period 1988-2013, varied considerably across cancer sites. Only 5 drugs for cancer of the eye, brain and central nervous system were registered, while 19 drugs for cancer of digestive organs were registered. The analysis will be based on aggregate data-longitudinal data on 15 cancer sites 3rather than patient-level data. Stukel et al. (2007) argue that comparisons of outcomes between patients treated and untreated in observational studies may be biased due to differences in patient prognosis between groups, often because of unobserved treatment selection biases. I believe that difference-in-differences estimates based on aggregate panel data are much less likely to be subject to unobserved treatment selection biases than estimates based on crosssectional patient-level data. 4 Moreover, the outcome measures that we analyze (premature mortality rates) are not subject to lead-time bias. 5 In Sect. 2, I describe an econometric model of premature cancer mortality. The data sources used to construct the data to estimate this model are described in Sect. 3. Empirical results are presented in Sect. 4. Key implications of the estimates are discussed in Sect. 5. Section 6 provides a summary and conclusions. 3 The 15 cancer sites are the 15 malignant neoplasm ICD-10 blocks defined by the World Health Organization. 4 Jalan and Ravallion (2001) argued that "aggregation to village level may well reduce measurement error or household-specific selection bias" (p. 10). 5 Survival time for cancer patients is usually measured from the day the cancer is diagnosed until the day they die. Patients are often diagnosed after they have signs and symptoms of cancer. If a screening test leads to a diagnosis before a patient has any symptoms, the patient's survival time is increased because the date of diagnosis is earlier. This increase in survival time makes it seem as though screened patients are living longer when that may not be happening. This is called lead-time bias. It could be that the only reason the survival time appears to be longer is that the date of diagnosis is earlier for the screened patients. But the screened patients may die at the same time they would have without the screening test. See National Cancer Institute (2015a).

Premature cancer mortality model
In his model of endogenous technological change, Romer (1990) hypothesized an aggregate production function such that an economy's output depends on the "stock of ideas" that have previously been developed, as well as on the economy's endowments of labor and capital. The premature mortality model that I will estimate may be considered a health production function, in which premature mortality is an inverse indicator of health output or outcomes, and the cumulative number of drugs approved is analogous to the stock of ideas. The first model will be of the following form: where, YPLL75 st = years of potential life lost before age 75 from cancer at site s per 100,000 population age 0-74 in year t (t = 2000, . . ., 2011); CUM_NCE i,t−k = d IND ds REGISTERED d,t−k = the number of new chemical entities (drugs) to treat cancer at site s that had been registered in Canada by the end of year t − k; IND ds = 1 if drug d is used to treat (indicated for) cancer at site s, 0 if drug d is not used to treat (indicated for) cancer at site s; REGISTERED d,t−k = 1 if drug d was registered in Canada by the end of year t − k, 0 if drug d was not registered in Canada by the end of year t − k; INC_RATE75 st = the average annual incidence rate of cancer at site s per 100,000 population age 0-74 in years t − 5, t − 4, . . ., t − 1 6 ; α i = a fixed effect for cancer at site s; δt = a fixed effect for year t.
Inclusion of year and cancer-site fixed effects controls for the overall decline in premature cancer mortality and for stable between-disease differences in premature mortality. A negative and significant estimate of β k in Eq. (1) would signify that diseases for which there was more pharmaceutical innovation had larger declines in premature mortality. The functional form of Eq. (1) has the property of diminishing marginal productivity: the absolute reduction in premature mortality declines with each successive increase in the number of drugs.
As illustrated by Fig. 4, the data exhibit heteroskedasticity-diseases with larger mean premature mortality rates had smaller (positive and negative) annual percentage fluctuations in YPLL75. Equation (1) will therefore be estimated by weighted least-squares, weighting by the mean premature mortality rate during 2000-2011 (( t YPLL75 it )/12). The standard errors of Eq. (1) will be clustered within cancer sites.
Although one would expect an increase in true cancer incidence to increase premature cancer mortality, cancer incidence rates are subject to measurement error, so one should not necessarily expect the coefficient on measured cancer incidence ( γ ) to be positive. Let I and I* represent measured and true cancer incidence, respectively. Then I = (I / I*) × I*, and log(I) = log(I / I*) + log(I*). Measured cancer incidence can increase for two reasons: an increase in true cancer incidence, or an increase in the ratio of measured incidence to true incidence. The latter could occur as a result of increasing quantity or quality of cancer screening. More and better cancer screening could lead to earlier diagnosis, which might reduce premature mortality. Therefore the effect on premature mortality of increases in I* and increases in (I / I*) may offset one another: the former is likely to increase premature mortality, but the latter may reduce it. For this reason, although controlling (in an unrestrictive manner) for measured incidence in the premature mortality model seems appropriate, we should not be surprised if we don't find a significant effect of measured incidence on premature mortality. (1)-the number of chemical substances previously commercialized to treat a disease-is not the theoretically ideal measure. Premature mortality is presumably more strongly related to the drugs actually used to treat a disease than it is to the drugs that could be used to treat the disease. A preferable measure is the mean vintage of drugs used to treat a disease, defined as VINTAGE st = d Q dst LAUNCH_YEAR d / d Q dst , where Q dst = the quantity of drug d used to treat cancer at site s in year t, and LAUNCH_YEAR d = the world launch year of drug d. 7 Unfortunately, measurement of VINTAGE st is infeasible: even though data on the total quantity of each drug in each year (Q d.t = s Q dst ) are available, many drugs are used to treat multiple diseases, 8 and there is no way to determine the quantity of drug d used to treat cancer at site s in year t. 9 However, it is shown in Appendix 1 of Lichtenberg (2014a) that there is a highly significant positive correlation across drug classes between changes in the (quantityweighted) vintage of drugs and changes in the number of chemical substances previously commercialized within the drug class.
Pharmaceutical innovation is not the only type of medical innovation that is likely to contribute to premature mortality. Other medical innovation, such as innovation in diagnostic imaging, surgical procedures, and medical devices, is also likely to affect premature mortality. Therefore, measures of these other types of medical innovation should be included in the Eq. (1). Unfortunately, longitudinal disease-level measures of non-pharmaceutical medical innovation are not available for Canada. But failure to control for non-pharmaceutical medical innovation is unlikely to bias estimates of the effect of pharmaceutical innovation on premature mortality, for two reasons. First, pharmaceuticals are more research-intensive than other types of medical care: in 2007, prescription drugs accounted for 10 % of U.S. health expenditure (Center for Medicare and Medicaid Services (2013, Table 2)), but more than half of U.S. funding for biomedical research came from pharmaceutical and biotechnology firms (Dorsey et al. 2010). Much of the rest came from the federal government (i.e. the NIH), and new drugs often build on upstream government research (Sampat and Lichtenberg 2011). The National Cancer Institute (2015b) says that it "has played an active role in the development of drugs for cancer treatment for 50 years… [and] that approximately one half of the chemotherapeutic drugs currently used by oncologists for cancer treatment were discovered and/or developed" at the National Cancer Institute.
Second, previous research based on U.S. data indicates that non-pharmaceutical medical innovation is not positively correlated across diseases with pharmaceutical innovation. In Appendix 2 of Lichtenberg (2014a), it is shown that, in the U.S. during the period 1997-2007, the rate of pharmaceutical innovation was not positively correlated across diseases with the rate of medical procedure innovation and may have been negatively correlated with the rate of diagnostic imaging innovation. Also, Lichtenberg (2014b) found that estimates of the effect of pharmaceutical innovation on U.S. cancer mortality rates were insensitive to the inclusion or exclusion of measures of non-pharmaceutical medical innovation. This suggests that failure to control for other medical innovation is unlikely to result in overestimation of the effect of pharmaceutical innovation on longevity growth.
In Eq. (1), premature mortality from cancer at site s in year t depends on the number of new chemical entities (drugs) to treat cancer at site s registered in Canada by the end of year t − k, i.e. there is a lag of k years. Equation (1) will be estimated for different values of k: k = 0, 5, 10, 15, 20, 25. 10 One would expect there to be a substantial lag because new drugs diffuse gradually-they won't be used widely until years after registration. Two kinds of evidence-"within molecule" and "between molecule"-support the gradual diffusion hypothesis. The first kind consists of estimates based on the π y parameters from the following equation: 9 Outpatient prescription drug claims usually don't show the indication of the drug prescribed. Claims for drugs administered by doctors and nurses (e.g. chemotherapy) often show the indication of the drug. In the US, 70 % of spending on anticancer drugs is for drugs covered under the medical benefit and infused or injected. However, t data on claims for drugs administered by doctors and nurses are not available for Canada. 10 A separate model is estimated for each value of k, rather than including multiple values (CUM_NCE i,t−1 , CUM_NCE i,t−2 , CUM_NCE i,t−3 ,…) in a single model because CUM_NCE is highly serially correlated (by construction), which would result in extremely high multicollinearity if multiple values were included.) where, SU my = the number of standard units 11 of molecule m sold in Canada y years after registration (y = 0, 1, …, 11); ρ m = a fixed effect for molecule m; π y = a fixed effect for age y The expression exp(π y − π 0 ) is a "relative utilization index": it is the mean ratio of the number of units of a molecule sold y years after registration to the number of units of the same molecule sold in the year that it was registered. Using annual data on the number of standard units of molecules sold in Canada during the period 1999-2010, I estimated Eq.
(2). Estimates of the "relative utilization index," based on data on 25 molecules used to treat cancer that were registered after 1998, are shown in Fig. 5. These estimates indicate that the number of units sold 10 years after registration is about ten times as great as the number of units sold one year after registration. Moreover, Fig. 5 provides a conservative estimate of the slope of the age-utilization profile, because there was zero utilization of many of these molecules in the first few years after they were registered. 12 Figure 6 provides "between-molecule" evidence of gradual diffusion; it shows data on the mean number of standard units of cancer drugs sold (in thousands) in Canada in 2010, by period of registration in Canada. Mean utilization in 2010 of drugs registered after 2000 is only 15 % as high as mean utilization of drugs registered during 1991-2000, and 17 % as high as mean utilization of drugs registered during 1981-1990.
The relatively low utilization of new drugs may be due to several factors. One is that the prices of old drugs (most of which are no longer patent-protected) are considerably lower than the prices of new, patent-protected drugs. A second factor may be that it takes time for physicians to become knowledgeable about new treatment options. A third potential factor is that new drugs may be targeted at smaller patient populations. Data from the U.S. Food 11 The number of standard 'dose' units sold is determined by taking the number of counting units sold divided by the standard unit factor which is the smallest common dose of a product form as defined by IMS HEALTH. For example, for oral solid forms the standard unit factor is one tablet or capsule whereas for syrup forms the standard unit factor is one teaspoon (5 ml) and injectable forms it is one ampoule or vial. Other measures of quantity, such as the number of patients using the drug, prescriptions for the drug, or defined daily doses of the drug, are not available. 12 Since the dependent variable of Eq. (2) is logarithmic, observations for which SU my = 0 had to be excluded.  (2015) indicate that drugs approved by the FDA since 2000 were twice as likely to include pharmacogenomic information in their labeling as drugs approved before 2000. A fourth potential factor is that older drugs are more likely to have supplemental indications, i.e. indications approved after the drug was initially launched, than new drugs. 13 The measure of pharmaceutical innovation, CUM_NCE s,t−k = d IND ds REGISTERED d,t−k , is based on whether drug d had an indication for cancer at site s at the end of 2011. One would prefer to base the measure on whether drug d had an indication for cancer at site s at the end of year t-k. FDA data indicate that about one in four new molecular entities has supplemental indications, i.e. indications approved after the drug was initially registered. 14 In Eq. (1), the measure of premature mortality is the number of years of potential life lost before age 75. This is the age threshold used in Statistics Canada's key socioeconomic database (CANSIM). Other authorities use different age thresholds; the CDC (2013) provides estimates of YPLL before ages 65, 70, 75, 80, and 85. To assess the robustness of my results, I will estimate models similar to Eq. (1), using age thresholds 65 and 55 as well as 75.
Chemical substances are divided into different groups according to the organ or system on which they act and their therapeutic, pharmacological and chemical properties. In the Anatomical Therapeutic Chemical (ATC) classification system developed by the World Health Organization Collaborating Centre for Drug Statistics Methodology, drugs are classified in groups at five different levels. The highest (1s) level is the "anatomical main group" 13 The measure of pharmaceutical innovation, CUM_NCE s,t−k = d IND ds REGISTERED d,t−k , is based on whether drug d had an indication for cancer at site s at the end of 2011. One would prefer to base the measure on whether drug d had an indication for cancer at site s at the end of year t -k. Data in the U.S. FDA's Drugs@FDA data files indicate that about one in four new molecular entities has supplemental indications, i.e. indications approved after the drug was initially launched. 14 Source: Drugs@FDA Data Files. level; there are 14 anatomical main groups. The 2nd, 3rd, 4th, and 5th levels are "therapeutic subgroup," "pharmacological subgroup," "chemical subgroup," and "chemical substance," respectively. 15 Premature mortality from a disease may depend on the number of chemical (or pharmacological) subgroups that have previously been developed to treat the disease rather than, or in addition to, the number of chemical substances (drugs) that have previously been developed to treat the disease. This will be investigated by estimating versions of Eq. (1) in which CUM_SUBGROUP s,t−k is included in addition to or instead of CUM_NCE s,t−k , where Drug indications (IND). Data on drug indications were obtained from Thériaque, a database of official, regulatory, and bibliographic information on all drugs available in France, intended for health professionals. This database is produced by the Centre National Hospi-15 For example, the five levels associated with the chemical subgroup "nitrogen mustard analogues" are: Nitrogen mustard analogues L01AA01 cyclophosphamide L01AA02 chlorambucil L01AA03 melphalan L01AA05 chlormethine L01AA06 ifosfamide L01AA07 trofosfamide L01AA08 prednimustine L01AA09 bendamustine

Empirical results
Estimates of the β k parameters from Eq. (1) and similar equations are shown in Table 2 and plotted (on an inverted scale) in Fig. 7. Each estimate was obtained from a separate model. All equations include a cancer incidence measure (e.g. ln(INC_RATE75 st )), cancer-site fixed effects and year fixed effects. To conserve space and simplify the presentation, estimates of the cancer incidence coefficient ( γ ) are not included in Table 2. 18 None of the estimates of this coefficient were statistically significant, and controlling for cancer incidence had very little effect on the estimates of β k . As discussed earlier, this may be due to offsetting effects of increases in I* and increases in (I / I*) on premature mortality. In part A of Table 2 and Fig. 7, the age threshold for calculating premature mortality is 75 years, i.e. the dependent variable is years of potential life lost before age 75. In model 1, the lag (k) from drug registrations to premature mortality equals zero, i.e. we are examining the effect of the cumulative number of drugs registered by the end of year t on premature mortality in year t. The estimate of β 0 is not statistically significant. In model 2, the lag is 5 years; the estimate of β 5 is also statistically insignificant. In models 3-6, the lags are 10, 15, 20, and 25 years, respectively. All of these coefficients are negative and highly statistically significant (p value <0.0001), indicating that premature mortality before age 75 is significantly inversely related to the cumulative number of drugs registered at least 10 years earlier. The estimate of β 15 is the most statistically significant, and the magnitude of the point estimate of β 15 is 66 % larger than the magnitude of the point estimate of β 10 . Since, as discussed earlier, mean utilization of drugs that have been marketed for less than 10 years is only one-sixth as great as mean utilization of drugs that have been marketed for at least a decade, it is not surprising that premature mortality is strongly inversely related only to the cumulative number of drugs that had been registered at least ten years earlier.
In parts B and C of Table 2 and Fig. 7, the age thresholds for calculating premature mortality are 65 years and 55 years, respectively. The estimates based on these age thresholds are very similar to the estimates based on the age threshold of 75 years: premature mortality before age 65 and 55 is strongly inversely related only to the cumulative number of drugs that had been registered at least ten years earlier. Figure 8 shows a bubble plot of the long-run (2000-2011) log change in YPLL before age 75 [ln(YPLL75 s,2011 )− ln(YPLL75 s,2000 )] against the long-run change in the cumulative number of drugs registered 15 years earlier [CUM_NCE s,1996− CUM_NCE s,1985 ], i.e. the number of drugs registered during the period 1985-1996. The bubble size is proportional to the mean premature mortality rate during 2000-2011 (( t YPLL75 it ) / 12). This figure confirms the finding from model 4 in Table 2 of a highly significant inverse relationship. The point estimate of β 15 from the long-difference model (β 15 = −0.0247, t value = −4.42; p value = 0.0010) is similar to the point estimate of β 15 from model 4 in Table 2. Figure 8 reveals that the largest number of drugs that were launched during 1985-1996 were for cancers of lymphoid, haematopoietic and related tissue (ICD-10 block C81-C96), which include leukemia. In principle, it is possible that excluding this ICD-10 block could have a substantial effect on the estimates, although it does not appear from Fig. 8 to be an outlier. As shown in Appendix Table 5, when ICD-10 block C81-C96 is excluded from the sample, estimates of Eq. (1) are very similar to the estimates when it is included.
As discussed above, the hypothesis that premature mortality from a disease depends on the number of chemical (or pharmacological) subgroups that have previously been developed to treat the disease rather than, or in addition to, the number of chemical substances (drugs) that have previously been developed to treat the disease can be tested by estimating (1) in which CUM_SUBGROUP s,t−k is included instead of, or in addition to, CUM_NCE s,t−k . Table 3 provides estimates of models suitable for testing this hypothesis. Estimates from three different models are presented there. In all three models, the dependent variable is ln(YPLL75 st ) and k = 15. The first model shown is the same as model 4 in and CUM_SUBGROUP s,t−15 . Controlling for the cumulative number of drugs, the cumulative number of chemical subgroups is not statistically significant. These estimates suggest that drugs (chemical substances) within the same class (chemical subgroup) are not "therapeutically equivalent," 19 i.e. they do not have essentially the same effect in the treatment of a disease or condition.

Discussion
During the period 2000-2011, the premature (before age 75) cancer mortality rate (the number of years of potential life lost due to cancer before age 75 per 100,000 population age 0-74) declined by about 9 %. The estimates of model 4 imply that, in the absence of pharmaceutical innovation during the period 1985-1996, the premature cancer mortality rate would have increased about 12 % during the period 2000-2011. 20, 21 As shown in Fig. 9, the premature mortality rate would have been 1788, rather than its actual value of 1459. In 2011, the population age 0-74 was about 32.1 million (or 321 hundred thousand), so the estimates of model 4 imply that pharmaceutical innovation during the period 1985-1996 reduced the number of years of potential life lost to cancer before age 75 in 2011 by 105,366 (= 321 * (1788 -1459)).
The premature (before age 75) cancer mortality rate is the number of years of potential life lost due to cancer before age 75 per 100,000 population age 0-74. This reduction in premature mortality is an estimate of the benefit to Canadians below age 75 in 2011 of pharmaceutical innovation during the period 1985-1996. Now I will calculate an estimate of the (social) cost of this innovation. As shown in Appendix Table 1 in supplementary matrial, 40 drugs that are used to treat cancer were registered during the period 1985-1996. Data from IMS Health indicate that in 2010, 22 expenditure on products containing these molecules was 409 million USD, which is about 1.9 % of total Canadian drug expenditure (21.6 billion USD). About 70 % of cancer patients were diagnosed before the age of 75, so it seems reasonable to assume that 288 million USD (= 70 % * 409 million USD) was spent on these drugs for cancer patients below the age of 75. 23 This implies that the cost per life-year before age 75 gained from previous pharmaceutical innovation was 2730 USD (= 288 million USD / 105,366 life-years).
Presumably most of the drugs registered during the period 1985-1996 were off-patent by 2010, so these cost estimates reflect prices of generic drugs. Law (2013) argues that Canadian generic drug prices have traditionally been set using a percentage of the equivalent brandname price as a ceiling, and that typically, these percentages ranged between 60 and 70 % of the brand price. 24 This suggests that if these drugs had been sold at branded rather than generic prices, the cost per life-year gained would have been between 3900 (= 2730 / 60 %) and 4550 (= 2730 / 70 %) USD. However, the ratio of generic price to branded price may be significantly lower for cancer drugs (which are often infused or injected) than it is for other drugs (which are primarily administered orally). For example, when imatinib, which is used to treat a set of leukemias, went generic in Canada in 2013, the generic drug price was approximately 25 % of the branded price. If the generic/branded price ratio were 25 %, and if these drugs had been sold at branded rather than generic prices, the cost per life-year gained would have been 10,920 (= 2730 / 25 %) USD. Hirth et al. (2000) performed a search of the value-of-life literature and identified 41 estimates of the value of life from 37 articles. 25 From estimates of the value of life, they calculated estimates of the value (in 1997 dollars) of a quality-adjusted life-year (QALY). 26 Four types of methods were used to produce those estimates: revealed preference/job risk, contingent valuation, revealed preference/non-occupational safety, and human capital. Median implied values (in 1997 and2011 dollars 27 ) of a QALY estimated in those studies are shown in the following My estimate of the cost per life-year before age 75 gained from previous pharmaceutical innovation is well below even the lowest estimates of the value of a life-year saved.

Summary and conclusions
The premature cancer mortality rate has been declining in Canada, and there has been considerable variation in the rate of decline across cancer sites. I analyzed the effect that pharmaceutical innovation has had on premature cancer mortality in Canada during the period 2000-2011, by investigating whether the cancer sites that experienced more pharmaceutical innovation had larger declines in the premature mortality rate, controlling for changes in the incidence rate.
The study is subject to several limitations. First, the measures of pharmaceutical innovation that were used were based only on labeled indications, but the National Cancer Institute (2015d) says that "off-label use of drugs is very common in cancer treatment." Second, it was not possible to measure or control for non-pharmaceutical medical innovation. Third, the outcome measures used were life-years gained, not quality-adjusted life-years gained.
The estimates indicated that premature mortality before age 75 is significantly inversely related to the cumulative number of drugs registered at least 10 years earlier. Since mean utilization of drugs that have been marketed for less than 10 years is only one-sixth as great as mean utilization of drugs that have been marketed for at least a decade, it is not surprising that premature mortality is strongly inversely related only to the cumulative number of drugs 25 Twenty-eight of the reviewed articles used U.S. data; the remaining articles used data from the U.K. (4), Canada (3), France (1), and Denmark (1). National origin did not significantly affect the values. 26 Lichtenberg (2009) demonstrated that, although the health of cancer patients is less than perfect, the number of QALYs gained from pharmaceutical innovation could be either greater than or less than the number of life-years gained. 27 The U.S. Consumer Price Index increased by 40 % between 1997 and 2011. that had been registered at least ten years earlier. Premature mortality before age 65 and 55 is also strongly inversely related to the cumulative number of drugs that had been registered at least ten years earlier. None of the estimates of the effect of incidence on mortality were statistically significant.
Controlling for the cumulative number of drugs, the cumulative number of chemical subgroups does not have a statistically significant effect on premature mortality. This suggests that drugs (chemical substances) within the same class (chemical subgroup) are not therapeutically equivalent.
During the period 2000-2011, the premature (before age 75) cancer mortality rate declined by about 9 %. The estimates imply that, in the absence of pharmaceutical innovation during the period 1985-1996, the premature cancer mortality rate would have increased about 12 % during the period 2000-2011. A substantial decline in the "competing risk" of death from cardiovascular disease could account for this. The estimates imply that pharmaceutical innovation during the period 1985-1996 reduced the number of years of potential life lost to cancer before age 75 in 2011 by 105,366.
The cost per life-year before age 75 gained from previous pharmaceutical innovation is estimated to have been 2730 USD. Most of the previously-registered drugs were off-patent by 2011, but evidence suggests that, even if these drugs had been sold at branded rather than generic prices, the cost per life-year gained would have been below 11,000 USD, a figure well below even the lowest estimates of the value of a life-year gained.
The largest reductions in premature mortality occur at least a decade after drugs are registered, when their utilization increases significantly. This suggests that, if Canada is to obtain substantial additional reductions in premature cancer mortality in the future (a decade or more from now) at a modest cost, pharmaceutical innovation (registration of new drugs) is needed today.

<0.0001
Each estimate was obtained from a separate model. All equations include a cancer incidence measure (e.g. ln(INC_RATE75 st )), cancer-site fixed effects and year fixed effects.
Standard errors are clustered within cancer sites