Tobacco Documents Online

568 Statement of -Eleanor J. tdacdonald Professor .°aneritus of Epidemiology Department of Cancer Prevention L'niversity of Texas System Cancer Center M. D. Anderson Hospital and Tumor Institute, Houston, Texas Mv name is Eleanor Macdonald, Professor Emeritus of Epidemiology at the University of Texas System Cancer Center, M.D. anderson Hospital and Tumor Institute, Houston. I have worked in the field of epidemiology for over 40 years in three state programs in Massachusetts, Connecticut and Texas; in fact, I established the first state cancer registry in the U. S. Prcm 1948 through 1974, I chaired the Department of Epidemiology at the University of Texas System Cancer Center. I have authored or coauthored about 150 publications, most of which deal wi"^h different phases of cancer epidemiology, and I remain actively involved in a number of ongoing research studies in epidemic'ogy. I an also the editor of the Epidemiology, Statistics and Cancer Control section of the Yearbook of Cancer. I share the concerns of this Committee regarding cancer morbidi`y and mortality, and Z encourage legitimate efforts to control this disease. However, as a scientist who has dedicated her professional career to a careful study of cancer epidemi- ology, I an appalled at the belief implicit in H.R. •495?, H.R. 5653, and S. 1929 that Congress can legislate scientific _-_:. _* urge this augu<_ -.c1-d putting itself in These bills a. :::d purpose because of of lung cancer. .->>es of lung cancer < --=::er science or gover- - o-4 identifying the r causation that H ='P-e, the depth of -_zing the effects -'s;.rv and environmen One of the r• ='_::dings regard of the prim --rents are based, i. _as ,,,,nducted iarc -'ryAG =:ze data of - s -:cre aware the = an most laymer -.~ncress must u. -, and no amount -••=^ animals d

=aftr J. Macdona2a -s of tpidenioloqy :aact= Prev.ntioa Systea Canc.r Center .aor Znststute. 1/oMstaya*-@4"* 1cdonald, Profs uor tw.!lts4 it :f Texas System Canc.r dMa%wt;_ :r Institute, youston. I Me+a .oqy for over 40 years ta tait" Connecticut ar.d Texasr sa :a". aacer registry in the i:. f. tivo e Department of Epid.eso.>qy at 'ancer Center. I have autnarsd ct )ns, most of which d.a: v:tL emiology, and i :ema:n research studies in ep:d.msc:oq;. demioicgy, Statist:cs and Cancrt ^f Cancer. f this Committee regardinq canCK encourage legitimate efforts t* as a scientist who has ded:cat.d reful study of cancer epideai- !lief implicit in H.Q."i95'. gress can legislate scientif-c 569 _a augus*_ body of we1'_-intentio.^.ed lec'_s_ators to =aelf in such an untenable ncsition. •se bills are a misdirection of governmental energy ruse of their narrow focus on smoking as t`e ma;or .ancer. We do not, in fact, know the cause or cancer and it is not in the best interest of or government to take a simplistic apprcach to the .__-7ing them. There are obviously many factors in that we have not yet begun to understand. For o*_h of our uncertainty is still far _co great _ffects of nutrition, work history, sty:e, ... _. cnment. = the main reasons that I disagree :ri _ n tzese regarding smoking and cancer stems fr_m my ^e primary type of evidence on whic: these :ased, i.e., epidemiolegical. As an epide=ic'-cgist ._-ed large-scale oopulaticn studies, and w=o ^as ata of vast numbers of researc:^, reocr_s, : a.:. :•.+are than many scientists -- and certa'_n:-, more :t _ay:~en -- of the limitations of .,.._s =:ce of ::.ius_ understand that epidemiology l .s n..c an exact a:^ount of wishing can make it so. We ca.. nct t.._.. .r._.ma1s di_ected at will toward various = -csur=_s. (2)

570 We cannot send all death certificates to one infallible pathologist for correction and confirmation. We cannot force all individuals supplying the raw data to epidemiologists, or even all epidemiologists themselves, to demand the type of precision necessary in data gathering and review to ir.sure that the cuali*_y of the data meets high scientific standards. Consequent':y, iudgments about chronic disease causation based on epidemiological findings must be highly tentative; they do not enjoy the absolute certainty implied in the findings of these three bills. :^crtality data are frequently used by eoidemiologists to study disease trends, often wit!:cut _°u'_1 awareness of inherent weaknesses in these data. :?ortal'-tl rates are cbtained from information on death certificates furnis'r.id by physicians or local health officials. Although I am certain that every effort is generally made to fill these out conscientiousi_, errors do occur -- in diagnosis as well as in recording of the data. In addition, other excgenous influences can cause pro- found fluctuations in mortality data over the years that can easily be misinterpreted as changing disease trends. For example, the introductions of each of the seven revisioiis of the International Lists of Diseases and Causes of Death (ICD) have had effects, generall_v improvements, on the classification of (3) __seases. :Bi_: each _e _isease trends has --een -.,rted. Sadly enough, = ain the classificati: _-oerfect. The possibili_ _zout a rise or fall ~- __..estigators question :_atis*_ics on causes of _=ace _^e patterns of ; -.•.:s t =emer.wer that the : _..=_.,iuced a classi'icat =:.~m metastatic cancer _..scecified as to orima ._s a sharp reduction i: --._ _ :en a rise in both -:::c d'_d it take for th =--..rately did physicia -oondar•f? When epidem_ -ncer trends, do they -_cor.darv cancers? We c -estions. ~ o77, t)----37

571 .s to one -..fallib'_e __ion. we cannot force all epidemiologists, or even and tie type of precision to insure that the cuali`y andards. Consecuent_i, _usation based on _1 tentative; they do not _n the findings of these ! used by eoidemiologists full awareness of inherent _ates are obtained fren iish=d by physicians or certain that every effort :nscientiously, errors co ording of the data. ?n_°'-uences can cause oro- :)ver the years that can disease trends. For :he seven revisions of the uses of Death (ICD) have n the classification of ..'_seases. ,vi_-" each __vis'_cn, however, _.__ _..~_ t: z disease trends has been bro;cen and in sc:^e ..3ses serious_v dis- torted. Sad'_v enough, even thcuch -_ has taken us centuries to attain the classificatior, system that we now have, _t is still i:^Derfect. '°he possibil_t7 of shiiting classi°ications ..ringing about a rise or fall in certain causes of death makes scme '_nvestigators question the value of any effort to compi'_e statistics on causes of death. For esanple, anycne atter..pting to trace _:e patterns of lung cancer during ,.he t_we.^.tieth cen tury must renember that the 1948 ICD revision, '_rst applied _.. 1949, _..z-ccuced a classi'ication secarat'_ng riTar7 cancer of the lung from metastatic cancer of _^e lung (ccded _hereafter as "~nsceci='_ed as to prinary or secondar_v). For t:oo ,ears, there was a s'r,arp reduction in the repcrted primary _,.nc cancer rate, and then a rise in bcth the p i:-ary and unspec'_'ied rates. 3ow lonc d'_d it take for the new classification to catch on? iow accurate'_p dia :hvsicians distinguish between ,-_:nary and secondar. ? svhen epidemiolcgists study z:went__.... __..tur_ '_,.-g cancer :_ends, do t:^.ev include or eSciude the unsneci_ied and secor.darv cancers? t.e cannot be sure about the answers t., _: ese cuestions. (4) -u77 o-s2-.,,

572 The 1965 ICD revision, acco*_ed in 1968, seriousl compl'_cated the situation by deleting the unspecified catego r. This revision effectively combined primary and unspecified lung cancer, thereby removing a needed safeguard for accuracy. Primary and secondary lung cancer are separate disease entities, with quite possibl_v distinctly different causes. Since ten ger- cent of all cancers spread to the lung, and since, for nearlv twenty years, more deaths were coded in the unspecified than in the primary category, the combination of those two categories seriously confused the lung cancer data. The damage done to epidemiological investigations of lung cancer by the 1965 classi- '_ication chanae cannot be underestimated. In the ninth revision (i980), greater confusion was added by placing the bronchus and _ung unspecified as the ninth subdivision of ari:iary lung ca ncer (162.9). Death certificates seldcm list more than a three-diait number, such as 162. t1e simply cannot separate cut °rom the mortali ty data the specific information we need to study '_una cancer scientificallv. Another aspect of the problem of changing disease classifications can be found b_v comparing trends in stomac^h and lung cancer. In the first half of this century, the phrase "question of gastric cancer" was used .._eauentl_i on death certi- ficates when a cachectic oatient died within davs or hours of first summoning a?hysician. It is fascinating to see that in states, within the repor 3ud.:enly as the uns .::anges may in par` from the difficult _'f:cult to diagnc Besides :^erail lung cance _,:crted lung canc -,_ years, those : cancer focus• - _.ers:iv tended t -.c wever, more att. __cluse of report-- _i ~ that increaE: this inter -=ser•raticns abcu zonclusions d- ?irst c ccking prevalenc =sZ _..formation yet the

573 _.. 1363, seriousl;~ the ..nsaeci:ied category. _mary and unsneci=ied lung afeguard for accuracy. seoarate disease entities, .nt causes. Since ten per- :g, and since, for nearly ia the unspecified than in : of those two categories ata. The damage done to : cancer bv the 1965 classi- ::ed. in the ninth revision v placing the bronchus and sion of primary lung cancer ist more than a three-digit :ot separate out from the '_on we need to study lung olem of changing disease 3ring trends in stomach and this century, the phrase : frequently on death certi- d within days or hours of fascinatir.g to see that in 47 states, wi_aiz two years of the ir.troduction of the 1348 _evision, the reported gastric cancer mort.ai:`_ rates dreo_'.ed as suddenl_v as the unspecified lung cancer ones rose. Thus, the two changes may in part represent a shift of cuestionable diagnosis from the difficult to diagnose stomach cancer to the ecually difficult to diagnose lung cancer classl=lcat:on. Besides studying anomalies that have developed in overall lung cancer trends, I have also examined trends in the reported lung cancer mortality rates for women in this century. ?or years, t:ose individuals who believed that smoking caused _ung cancer `ocused almost solely cn male lung cancer, and generaliy tended to ignore female lung cancer. Irn recent years, however, more attention has been given to female lung cancer because of reoorts of an increase in female lu.^.g cancer and the claim that increased cigarette smoking was responsibie. I do not ;ccea* this internretation because I have serious sc'_entific reservaticns abcut both the accurscy of the underlving data and the conclusions drawn frcm it. First of all, we simply do not have reliable data on smokina oreva:ence in women -- or in men, for that matter. The best information we have is from the National Health Survey =eccrts, yet the 1979 reoort, for exampie, has a standard error (6)

574 of 30 percent. That percentage of error is too high to prcvi4-e a basis for drawina definitive conclusions. Secondly, I have doubts about whether lung cancer in women has actually increased dramatically. As I indicated earlier, national mortality data may not be reliable. _f we assume for the sake of argument, however, that they are _eiiable, We can still question the argument that female lung cancer has increased sharply in recent years. Although the age-ad'_usted J.S. death rates from respiratory cancer in white females show a consistent increase from 1953 through 1975, the rate of t::is increase has varied over this period. A smoothed year-to-year variation in the rate of increase -- the slope -- indicates that a sharp rise began around 1960, but leveled off around 1970. Althougn the rate of increase has begun to rise again, it is not as steep an ascent as in the 1960's. This is important because the slope gives a predictive picture of the overall mortality trend. And the slope that I see for female lung cancer suggests a stabilization. It :-iay even nredict a decline in lunc cancer mcrtalitv in the coming years. Other factors must also be considered in an anal.rsis of the apparent increase in female lung cancer. One such =actor, inaccurate repor`ina, appears _o have played apredominant _.,__. (7) analvsis of b z:~-ests that lu: _s being ove= Jnderr( =..r=_e clinical 4.. -_ncer, 2) inade .-acnosis of lur __saases. First _eve that wcr. _..c_ , and thac =_ond, phvsicia -,cause thev sir. -._ resence of -._se li:litation _..~_ resoirator ~:ey may had lung

'. error is too high to prcvide a _•lsions. about whether lung cancer in :matically. As I indicated mav not be reliable. If we .cwever, that they are reliable, t that female lung cancer has ,. Although the age-adjusted cancer in white females show a :ough 1975, the rate of this iod. A smoothed vear-to-vear -- the slcoe -- indicates that :ut leveled of= around 1970. begun to rise again, it is not s. e the slope gives a predictive rend. And the slope that I see stabilization. It mav even ~rtalitp in the coming years. ce considered in an analysis of :r.g cancer. Cne such factor, ave played a predominant role. 575 xy analysis of both reporting techniques and _.._ r.cer:_~i~c d_• _a suggests that lung cancer in women was underrepor_ed _.. _.._ ~as= and is being overreported now. Underreporting may have occurred pri^:ari'_: because of three c'_inical factors: 1) less ciinicai interest _.n female :ung cancer, 2) inadequate diagnostic tools, and 3) clinicai mis- iacnosis of lung cancer as tuberculosis or other _espiratory diseases. First of all, for many years, clinicians tended believe that women were less likely than men to develcp luaa cancer, and that belief could have affected their diaar.eses. Second, physicians were less capable of detecting lung cancer, because they simply did not have adequate means c= d>scoveri::c the presence of a cancer. Third, mainly as a consequence e' _i:ese limitations, physicians may have confused lung cancer wioh ct:er resniratory diseases, particu'_ar'_v tuberculos'_s. 7._ ot:zer words, t`.:e,7 may have diagnosed tuberculosis N:en in fact t~ie catient had lung cancer. The introduction of antibiotics and other drugs led _c a shar= decline in deaths from tuberculosis and other _..'ect'_eus diseases. With isoniazide and these other drugs at their dis- gosal, physicians could suddenly separate the luna cancer tat'_ents from the tuberculosis patients, because r's.e T3 patients .anerall. i:aproved after drug therapy whereas _:e cancer patier.ts (8)

576 did not. The abrupt drop in tuberculosis rates _f___ the zntrcduction of isoniazide in 1952, and t:ze nea- simultaneous increase in reported lung cancer rates, provide substantial suppcrt for my belief that these diseases were o:ten confused clinicallv. Therefore, as these druas became _ncreasing'_y available, underreporting of female lung cancer became less common. IInfortunate'_y, there is evidence which suggests that today lung cancer in women is overreported. This may occur in one of two wavs: (1) clinicians are not always able to dis- tinguish between primary and secondary lung cancer. As discussed earlier, the distinction is of obvious '_mportance for a factual cour.t of pr'_marv lung cancer cases. T.`.e, only truly effective way to determine whether lunc cancer is orimarv or secor.darv is through autopsv, and few cases of cancer deaths in this countr,l are autopsied. In general, the autoosy rate is about 12 percent. When one considers that this percentage includes all the accidental and viclent deaths that must be autopsied, or.e understands ]IIst how 1Cw the autCpsV rate f.._ lung cancer may be. This issue becomes even more crucial because pr'_-:ary lung cancer is one of the more difficult diacnoses to establish ci'_nicallv. (2) The current ICD classification combines primar^'•and unspecified lung cancer into one category. Even i'_ the cl_n'_c'_an could cerrectl_v distinguish primary from unspecif_ed lung cancer, (9) z:e effort would be . hcth orinarv and unsr reports that have des, Therefore, c _n disease nomenclatur the epidemiologists' : reoorted trends in 1ur. =rue incidence of the Further sup __..___'_cat'-on of ciga =_..cer stems from my c :~re, a study of 1'. .c,.en in El Paso, we f =%anish surnamed, 64 r --me percentage of tht cancer was half t '-.a= we had to rule - ---s _aad , t he most i MF -,en's lung cancer wa adobe houses. Tyee =---_ ventilated stru S-d the women to ;c

577 .csis mcrtall=•/ rates a=ter and near'_y cancer rates, prov:de = these diseases were often these drucs became _ of 'ema'_e lung cancer 3ence which succesus that orted. This may occur in ^ot always able to dis- lung cancer. As discussed imoortance for a -factual ..ey on1_, truly e:=ect'_ve way imary or secondary is :er deaths in this coun*_ry ~ rate is about 12 percent. :age includes all the ;ust be aL`toDslea, ore ate :cr lung cancer aav be. ecause primary lung cancer 3 to establish ci'_nicall_v. combines primar•."• and )r.),. Even if the cii.^.'_cian :m unspeci:ied lung cancer, t:e effort would be negated by classification met`.cds. Thus, both prinary and unspeciiied lung cancers are combined a .n the reports that have described increases in lung cancer in women. Therefore, changes '_n diagnostic techniques and chances iz disease ncmenclature are very important because, speaking from the epidemiologists' point of view, we cannct be sure that the reported trends in lung cancer among women accurately reflect the true incidence or the disease. Further support for my skepticism regarding an _denti`ication of cigarette smoking as the primary cause of lung cancer stems from my own research _nvestigations of this disease. _•n one, a study of lung cancer in Saanish surnamed and Anglo women in El Paso, we =ound a high incidence of the disease in the Spanish surnamed, 64 percent of whom were smokers. Although the same percentage of the Anclo women also smoked, their rate of lung cancer was half that of the Spanish surnamed. tve concluded that we had to rule out smoking as tae significant factor. lnstead, the most important factor i.^n the Spanish surnamed women's lung cancer was their residence from birth to adul=:ced in adobe houses. These are solidly built, nearly airtight and poorly ventilated structures heated mostly by wood tires, which exposed the women tc known carcinogens.