Tobacco Documents Online

CANCER UNDEFEATED CANCEK UNDEFEATED JOHN C. BAILAR III, M.D., PH.D., AND HEATHER L. GORNIK, M.H.S. ABSTRACT Background Despite decades of basic and clinical research and trialS of promising new therapies, can- cer remains a major cause of morbidity and mortal- ity. We assessed overall progress against cancer in the United States from 1970 through 1994 by analyz- ing changes in age-adjusted mortality rates. Methods We obtained from the National Center for Health Statistics data on all deaths from cancer and from cancer at specific sites, as well as on deaths due to cancer according to age, race, and sex, for the years 1970 through 1994. We computed age-specific mor- tality rates and adjusted them to the age distribution of the U.S. population in 1990. J~sult~ Age-adjusted mortality due to ~ancer in 1994 (200.9 per 100,000 population) was 6.0 percent higher than the rate in 1970 (189.6 per 100,000). After decades of steady increases, the age-adjusted mop tality due to all malignant neoplasms plateaued, then decreased by 1.0 percent from 1991 to 1994. The de- cline in mortality due to cancer was greatest among black males and among persons under 55 years of age. Mortality among white males 55 o~ older has also declined recently. These trends reflect a combi- nation of changes in death rates from specific types of cancer, with important declines due to reduced cigarette smoking and improved screening and a mixture of increases and decreases in the incidence of types of cancer not closely related to tobacco use. Conclusions The war against cancer is far from over. Observed changes in mortality due to cancer primarily reflect changing incidence or early detec- tion. The effect of new treatments for cancer on mor- tality has been largely disappointing. The most prom- ising approach to the control of cancer is a national commitment to prevention, with a concomitant rebal- ancing of the focus and funding of research. (N Engl J Med 1997;336:1569-74.) ©1997, Massachusetts Medical Society. IN 1986, when one of us reported on trends in the incidence of cancer in the United States from 1950 through 1982,1 it was clear that some 40 years of cancer research, centered pri- marily on treatment, had failed to reverse a long, slow increase in mortality. Here we update that analysis through 1994. Our evaluation begins with 1970, both to provide some overlap with the previ- ous article and because passage of the National Can- cer Act of 1971 marked a critical increase in the magnitude and vigor of the nation's efforts in cancer research,z The 1986 report and follow-up articles~,s's were criticized,~s primarily on the grounds that research already completed had not yet been incorporated into practice and that new research findings were on the way. Critics also argued that data for all cancers combined are not meaningful and that the study of age-adjusted mortality rates is not appropriate when the rates in different age groups exhibit different trends, as they do for cancer. The Senate asked the National Cancer Institute to convene a committee to consider how to measure progress against cancer, and it published its report in 1990.9 The committee recommended that progress be assessed in three general areas: direct measures (mortality, incidence, and survival, including the quality of life ), portents of change ( such as reductions in tobacco use), and advances in knowledge that may have an effect in the future.9 Direct measures were taken to be central to the assessment of progress. The most basic measure of progress against cancer is age-adjusted mortality. The use of rates removes the effect of changes in the overall size of the pop- ulation. Adjustment for age further removes the ef- fect of changes in the age distribution of the popu- lation, and with it the effect of changing mortality from causes other than cancer. The use of mortality as the chief measure of progress against cancer, rath- er than incidence or survival, focuses attention on the outcome that is most reliably reported and is of greatest concern to the public: death. The use of rates for all types of cancer combined, though diffi- cult to interpret in biologic terms, usefully supple- ments site-specific rates because it prevents selective reporting of data to support particular views and min- imizes the effects of changes in the diagnosis and re- porting of specific ,types of cancer. Briefly summarized, the reason for not focusing on the reported incidence of cancer is that the scope and precision of diagnostic information, practices in screening and early detection, and criteria for re- From the Department of Health Studies~ University of Chicago, 5841 S. Maryland Ave., MC 2007, Chicago, IL 60637-1470, where reprint re- quests should be addressed to Dr. Bai|ar. Volume 336 Number 22 1569

The New England lournal of Medicine porting cancer have changed so much over time that trends in incidence are not reliable.t For example, the development and vigorous commercial promo- tion of the test for prostate-specific antigen occurred at the same time as a doubling of the reported inci- dence of cancer of the prostate between 1974 and 1990 (from 65.6 per 100,000 population to 131.8 per 100,000),*0 without visibly affecting mortality. Few knowledgeable observers believe that either the true frequency or the lethality of the disease has changed much. A similar but smaller trend has af- f~:cted rates of breast cancer, and there are reasons for concern about the incidence of other cancersJ Trends in survival rates are also suspect, because they are based on the same series of patients as inci- dence rates, and any inflation of incidence due to the inclusion of less malignant or nonmalignant diseases creates a spurious increase in case survival rates. METHODS Sources of Data Nuhabers of deaths according to year, age, race, sex, and cancer site were obtained fi'om the National Center for Health Sratis- ticsJ' Population data came from the Bureau of the Census and the National Center for Health Statisticsu.*3 (and Rosenberg H, Mortality Statistics Branch: personal communication). Other data were obtained from the National Cancer Institute.t0 Age-specific mortality rates, the building blocks of age-adjusted rates, are simple ratios of numbers of deaths to the size of the pop- ulation. The numerators are the numbers of deaths from a specific cancer or group of cancers among people in specific age ranges and, often, with specific demographic characteristics. The denom- inator is the corresponding U.S. population, as estimated by the Bureau of the Census. Data adjusted for age by the "direct" meth- od (which we use throughout) are weighted sums of these age- specific rates, with the weights determined by reference to some fixed population, such as the total U.S. population in the 1990 ceusus..4 For example, age adjustment of rates for each of the years from 1984 through 1994 to the 1990 standard entails the estima- tion of mortality as if the actual population in each of those years had the same age distribution as the 1990 U.S. population. If we want to examine recent changes in overall mortality due to cancer, the most appropriate refi:rence population for adjust- ment is one that falls within, or very close to, the period of study. Because we are tbcusing largely on ~vents in recent years, we have used the U.S. population as reflected in the 1990 census. When trends in different age groups diverge, the choice of a reference population can make a substantial difference in estimat- ed trends. For example, the population of the United States was much younger in 1940 than in 1990, and hence the use of the 1940 population as the reference group gives greater weight to mortality rates among younger persons, which have been declin- ing, whereas rates in older persons have been increasing. There- fore, the 1940 standard gives an unduly favorable picture of re- cent trends in mortality due to cancer; rates adjusted to the 1970 standard lie between those adjusted to 1940 and those adjusted to 1990. Data presented at a recent press conference by the Depart- ment of Health and Human Services and the American Cancer Society, and in a related publication, reported rates that were ad- justed to the 1970 and 1940 populations.*sat RESULTS Table 1 shows age-adiusted death rates for all ma- lignant neoplasms, year by year, since 1986. For the U.S. population as a whole, the long-sustained annu- al increase in mortality due to cancer ceased in about 1991. Between 1991, when the highest rate xvas re- ported, and 1994, the most recent year tbr which data are available, mortality decreased by 1.0 percent (from 203.0 to 200.9 per 100,000 population). This drop may well portend larger improvements to come. Even if rates turn upward again, the decline will sure- ly resume within the next few years as a result of re- ductions in smoking over recent decades. For historical perspective, U.S. cancer mortality rates, age adiusted to 1970 by the National Cancer Institute, increased by an estimated 0.3 percent an- nually from 1975 through 1993, as compared with an increase of 0.1 percent per year from 1950 through 197570 This accelerated increase in mortality due to cancer occurred despite the enlarged scope of cancer research since 1971. Figure 1 presents trends in mortality from all ma- lignant neoplasms since 1970, according to race and sex. After decades of rather steady increases in each demographic group, mortality rates plateaued or de- clined slightly in the 1990S, most notably in the black male population, among whom the recent downward trend follows years of rapidly increasing mortality. Figure 2 shows trends since 1970 for males and females in two broad age groups. The population under 55 years of age is much larger than the older population, whereas rates of mortality due to cancer are much higher among older pe.ople than in the younger age group. As a result, the smaller percent- age increase in mortality observed in the smaller, older group represents more deaths than the larger percentage decrease in the younger group. The in- terplay of these factors determines the population- wide rate, which has changed much more slowly than rates within these two broad age groups. Among older persons, both men and women, mortality due to cancer increased by 15 to 20 per- cent between 1970 and 1994, with a recent decline among older men. During the same period, mortal- ity due to cancer among people younger than 55 de- creased by about 25 percent for both sexes. The close parallels between the rates for males and fe- males in each age category seem coincidental, since the rates for the two sexes reflect distinct patterns of cancer sites. We turn now to some specific forms of cancer. MorTality due to breast cancer has increased by ap- proximately 10 percent since 1970 among women 55 years of age or older, with a recent plateau, but has decreased by almost 25 percent among younger women (Fig. 3). The recent and substantial increase in the use ofmammography among women over 50, for whom annual examination is known to be effec- tive, has not prevented this increase. These data sug- gest that a true increase in incidence may have been 1570 May 29, 1997

CANCER UNDEFEATED TASLS 1. RECENT T~.~S ~N MORTALIT'f DUE TO CANCER IN THE UNITED STATES.* Y~k'~ TOTAL ~ FEM~ death=/100,000 1'986 1'99.0 256.4 1'61.3 1,987 1.99.2 256.7 161.3 [988 199.8 256.8 1'62.3 1989 201,.6 258.4 1'64.1 1990 202.4 259.6 164.6 1991 203.0 259.3 165.7 1992 201.8 256.7 165.3 I993 202.1 256.5 165.7 1994 200.9 253.2 165.7 *The rates shown are numbers of deaths from all malignant neoplasms per 100,000 population. Rates have been adjusted for age, with standardization to the age distribution of the U.S. resident population in 1990. only partially offset by the effectiveness of screening. Although mammography before the age of 50 is controversial, these data suggest that declines in mortality were well established before mammogra- phy became widely used. Overall, the decrease among younger women and the increase among older wom- en have left population-wide mortality almost un- changed. For lung cancer, death rates for women 55 or old- er have increased to almost four times the 1970 rate, whereas rates among males younger than 55 have decreased slightly (Fig. 4). Rates for older men and younger women have risen since 1970, but with some recent downturn. These trends reflect delayed effects of changes in smoking habits that occurred decades ago. Figure 5 shows trends in mortality for addition- al types of cancer from 1970 through 1993. Age- adjusted rates for several important types of cancer declined steadily. The decrease in cancer of the stom- ach, observed worldwide over many decades, is not well understood, but it is largely or entirely a result .of decreasing incidence rather than earlier detection or improved therapy. The sharp decline fbr cancer of the cervix is also not fully explained but reflects a combination of reduced incidence and improvements in the detection of premalignant lesions by means of the Papanicolaou smear and their subsequent remov- al; earlier detection of invasive cervical neoplasms may also be important. Deaths from cancer of the uterus (including uterine neoplasms not specified as of the cervix) are primarily due to endometrial can- cers, but they include a small proportion of deaths from cervical cancer reported as nonspecific cancer of the uterus and a few malignant myometrial neo- plasms. Here, too, there has been a sustained de- 350" 250" 150" 19'70 Black males White males Total Black females ~'~; females Figure 1. MoRality from All Malignant Neoplasms, 1970 through 1994, in the Total U.S. Population and According to Race and Sex. The rates have been age-adjusted to the U.S. resident poPula- tion of 1990. 120 - 70' 19i70 Rgure 2. Mortality from All Malignant Neoplasms, 1970 through 1994, in the Total U.S. Population as a Percentage of the Rate in 1970, According to Age and Sex. The rates have been age-adjusted to the U.S. resident popula- tion of 1990. 115- Females >55 yr o "~'" All females ~ 105" ¢~ 95" o yr 75- ,,, 1970 19'75 ~9'80 1985 1~90 Figure 3. Mortality from Breast Cancer, 1970 "through 1993, in the Total U.S. Female Population as a Percentage of the Rate in 1970, According to Age. The rates have been age-adjusted to the U.S. female resident population of 19§0. Volume 336 Number 22 1571

The New England Journal of Medicine 400 300' 200. 100' 19~0 Females 955 yr ,~'" Males ~55 yr ~.~ ................... Males 0-~ yr ~ 1 8s Figure 4. Mortality from Cancer of the Trachea, Bronchus, or Lung, 1970 through 1993, in the Total U.S. Population as a Per- centage of the Rate in 1970, According to Age and Sex. The rates have been age-adjusted to the U.S. resident popula- tion of 1990. 30. ~Colorectal cancer 20- Lymphoma and "'*'~ other cancers ~'~-t-~ ~ "~'~'*', ~ Stomach cancer ~-'~'-~'*~'~-'-~=-,=-,~.=.~-=--.-~o Cancer of brain aa=-~.~...~ ~-----,;-.,-,.~ Cancer of uterus "~ "~--"~-~-e Cervical cancer ~ - -" .... '- ..... Melanoma O" 19 0 19'7s 9'so 19's 1 90 Figure 5. Mortality from Cancer at Selected Sites, 1970 through 1993, in the Total U.S. Population. The rates have been age-adjusted to the U.S. resident popula- tion of 1990. cline, though not as great as tbr cervical cancer, and at least a part of this improvement is due to earlier detection. Mortality from leukemia (all types and in all age groups) has also decreased. Deaths from colorectal cancer (including anal cancer) decreased substan- tially for reasons that are not entirely clear, but they may include earlier detection as well as a reduction in incidence J0 Improved treatment has contributed little. Small increases have been reported for malignant brain tumors and malignant melanoma, shosvn here since 1979, when the National Center for Health Statistics introduced a new format for reporting mor- tality data.~* Mortality from lymphomas and other lymphoid neoplasms (including Hodgkin's disease, non-Hodgldn's lymphoma, and multiple myeloma) increased by 1Z3 percent from 1970 to 1993, de- spite reductions in mortality from Hodgkin's disease alone.~0 Trends in mortality due to cancer among children require special comment. Death rates for each major category of childhood cancer have declined by about 50 percent since the 1970s (data not shown). The decline is continuing, and the percentage drop in the most recent 10-year period is slightly greater than that for the previous 10 years. To put this find- ing in perspective, however, cancer accounted for only 1699 deaths among children under 15 years of age in the United States in 1993, among a total of 529,904 deaths due to cancer in all age groups.** Even the complete elimination of deaths due to child- hood cancer would have little effect on the national death toll. DISCUSSION It is worth reviewing probable reasons for these changes in mortality due to cancer. Some declines are clearly a result of reduced incidence or earlier de- tection (cancer of the cervix, other cancers of" the uterus, and cancers of the colon, rectum, and stom- ach). Similarly, recent changes in mortality from lung cancer are certainly due to changes in smoking pat- terns over the past few decades. The smaller increas- es in mortality from melanoma and cancer of the brain, the prostate, and perhaps the breast (in older women) can hardly be due to a declinein the effective- ness of treatment; they must reflect rising incidence. Thus, the observed trends largely reflect changing in- cidence or earlier detection, rather than improved therapy. Despite numerous past claims that success was just around the comer, mortality due to cancer contin- ued to increase, until quite recently. The death rate in 1994 was 2.7 percent higher than in 1982, the last year covered in the 1986 paper,t but it is likely that the recent downturn will be confirmed and substan- tially extended as a result of improved prevention 1572 May 29, 1997

CANCER UNDEFEATED and earlier detection and, especially, past reductions in tobacco use. In 1986, we concluded that "some 35 years of in- tense effort focused largely on improving treatment must be judged a qualified failure."~ Now, with 12 more years of data and experience, we see little rea- son to change that conclusion, though this assess- ment must be tempered by the recognition of some areas of important progress. These include the much- improved outlook for children and young adults with cancer, which is entirely the result of improved treatment; better treatment for Hodgkin's disease; far better palliation of many kinds of advanced can- cer; a better understanding of cancer, which as a by- product has improved the medical management of nonmalignant immunologic, metabolic, and viral diseases, including the acquired immufiodeficiency syndrome; and great improvements in imaging tech- nology. Though these benefits must not be dis- counted, their effects on overall mortality due to cancer have been largely disappointing• The argument that rising incidence has just bal- anced rising case survival rates, so that mortality, is roughly constant, seems unlikely to be true but is ir- relevant anyway. However one analyzes and inter- prets the present data, the salient fact remains that age-adjusted rates of death due to cancer are now barely declining. Hopes for a substantial reduction in mortality by the year 2000 were clearly mis- placed.~7 The effect of primary, prevention (e.g., re- ductions in the prevalence of smoking) and second- ary prevention (e.g., the Papanicolaou smear) on mortality due to cancer indicates a pressing need for reevaluation of the dominant research strategies of the past 40 years, particularly the emphasis on im- proving treatments, and a redirection of effort to- ward prevention. Unfortunately, the means to prevent most cancers have not yet been elucidated, adequately tested, and shown to be effective and feasible• For example, we need to know more about how to help the smoker who wants to quit, and much of the evidence that diet is related to one third or more of cancers~s must be reduced to findings about specific dietary, com- ponents. The needed research on prevention may demand as much in time, effort, and resources as has already been invested in studies of treatment. We em- phatically do not propose that research on treatment be stopped; there should, however, be a substantial realignment of the balance between treatment and prevention, and in an age of limited resources this may well mean curtailing efforts focused on therapy. Prevention is much broader than the elimination of carcinogens. For example, recent progress in un- derstanding the roles of dietary modification, chemo- • prophylaxis (e.g., with retinoic acid and tamoxifen), and genetic predispositions to cancer (in order to reduce exposure to carcinogens and to increase sur- veillance with the goal of earlier detection) holds in- triguing promise for substantial reductions in mor- talit.v due to cancer, although much critical research remains to be done. Also part of "prevention" re- search is the investigation of risk factors for cancer in order to determine which factors can be modified and im, estigations in the behavioral sciences aimed at improving the application of findings relevant to prevention. The role of basic research is unclear, partly because what is called "basic" is highly sub- jective and can be rapidly redefined in response to threatened budget cuts. However, we support the expansion of basic-science research that is not so basic as to have no clear, direct, and specific link to prevention. Will we at some future time do better in the war against cancer? The present optimism about new therapeutic approaches ~ooted in molecular medi- cine may turn out to be justified, but the arguments are similar in tone and rhetoric to those of decades past about chemotherapy, tumor virology, immu- nology, and other approaches. In our view, prudence requires a skeptical view of the tacit assumption that marvelous new treatments for cancer are just waiting to be discovered. We, like others, earnestly hope that such discover- ies can and will be made, but it is now evident that the worldwide cancer research effort should un- dergo a substantial shift toward efforts to improve prevention. Will this shift mean that prevention re- search will ultimately succeed in the way that treat- ment research was expected to succeed? There is no guarantee that it will. The ultimate results may be as disappointing as those to date from treatment ef- forts, but it is time to find out. There are also questions of implementation. Pre- vention is likely to be more difficult and costly than treatment, whicl-; can be rather narrowly focused on persons in need during a limited time and can be provided without major changes in the ambient en- vironment, workplace, diet, or consumer products. Treatment, if it could be made to work, would ob- viously be much simpler. The public seems to understand the need for the shift in attitude and emphasis toward prevention. The evidence includes the large and continuing reduction in smoking, widespread individual efforts to change diet to prevent cancer, and the use of sunscreens to reduce exposure to sunlight. The government has had little role in these changes. However, to leave this matter entirely to the public is to risk faddism, on the one hand, and a turning aveay from orthodox ther- apy, on the other. Aside from overstatement of the decline in mor- tality due to cancer in the United States in recent years, the recent joint press conference~ held by the Department of Health and Human Services and the American Cancer Society ~vas notable for its pubiic Volume 336 Number 22 1573

The New England Journal of Medicine recognition of the importance of prevention in the effort to control cancer. According to Secretary. of Health and Human Services Donna Shalala, We must continue to work for the day when our children must turn to the history books .to learn about a disease called cancer.... It will take better research, better treatments, better detection, and most important, it will take better education .... From tobacco to poor diet to lack of reproductive screenings, we must give the Ameri- can people the in~brmation they need to prevent cancer and make the best choices with their lives.*s We hope that this statement, as well as the recent in- crease in support of prevention activities in the Na- tional Cancer Institute budget,.9 represents an early step in the commitment to prevention, rather than lip service obscuring blind faith in treatment-based approaches. The best of modern medicine has much to offer to virtually every patient with cancer, for palliation if not always for cure, and every patient should have access to the earliest possible diagnosis and the best possible treatment. The problem is the lack of substantial improvement over what treatment could already accomplish some decades ago. A national commitment to the prevention of cancer, largely re- placing reliance on hopes for universal cures, is now the way to go. Presented in part as the Ramazzini Lecture, given by Dr. Bailar on October 26, 1996, in Carpi, Italy, as part of the annual Ra- mazzini Days of the Collegium Ramazzini. We are indebted to the National Center for Health Statistics for supplying most of the data used in this study; to the National Cancer Institute and the Bureau of the Census for the remainder; and to Dr. Samuel Broder for kindly suggesting the title. REFERENCES 1. Bailar JC III, Smith EM. Progress against cancer? N Engl J Med 1q86~ 314:1226-32. ~Z. National Cancer Act, P. L. No. 99-158 (1971). 8. Bailar JC. Rethinking the war on cancer. Issues Sci Technol 1987: 4:16-21. 4. Ide~. Cancer in Canada: recent trends in mortality. Chronic Dis Can 1992;l$'~uppl:S2-58. §. Ide~. Deaths from all cancer: trends in sixteen countries. Ann N Y Acad Sci 1990;609:49-56. 8. Progress against cancer? N Engl l Med 1986~315:963-8. 7. More on progress against cancer. N Engl I Med 1987;316:752-4. 8. The war on cancer:, views ~?om the fixmt. Issues Sei Technol 1988; 4:14-6. ~. Extramural Committee to Assess Measures of Progress Against Cancer. Measurement of progress against cancer. J Natl Cancer lnst 1990;82: 825-35. 10, Ries LAG, Kosary CL, Hankey BF, Harrm A, Miller BA, Edwards eds. SEER cancer statistics revi~v, 1973-1993: tables and graphs. Bethes- da, Md.: National Cancer Institute (in press). (For preliminary, edition see < http://www-seer.ims.nci.nih.gov>. ) ~ L National Center for Health Statistics. Vital statistics of the United States, 1970-1994. Vol. 2. Mortality. Part A. Washington, D.C.: Govern- ment Printing Office, 1974-1996 (1993 and 1994 in press). ~o Bureau of the Census. Preliminary. estimates of the population of the United States, by. age, sex, and race: 1970-198L Current population re- ports. Series P-25. No. 917. Washington, D.C.: Go~'emment Printing lice, 1982. 1~. Idem. U.S. population estimates, by age, sex, race, and Hispanic origin: 1980-1991. Current population report~. Series P-2S. No. 109S. Washing- ton, D.C.: Government Printing Office, 1993. 1~. Breslow NE, Day NE. Statistical methods in cancer research. Vol. 2. The design and analysis of cohort studies. Lyon, France: International Agency for l~:search on Cancer, 1987. (IARC scientific publications no. 82.) ~§. Shalala DE. New cancer mortality rates. Washington, D.C.: Depart- ment of Health and Human Services, 1996 (press conferenceS. ~8. Cole R Rodu B. Declining cancer mortality in the United States. Can- cer 1996;78:2045-8. 1Z National Cancer Institute. Cancer control: objectives for the nation, 1985-2000. NCI monographs. No. 2. Washington, D.C.: Government Printing Office, 1986. (NIH publication no. 86-2880.) 18. Doll lk, Peto R. The causes of'cancer: quantitative estimates of avoid- able risks of cancer in the United States today. J Natl Cancer Inst 1981; 66:1191-308. "1~. Financial Management Branch. NCI Ikct book. Bethesda, Md.: Na- tional Cancer Institute, 1993-1996. 1574 May 29, 1997