Tobacco Documents Online

Lies, , damned lies an d Doctors, journalists and health food vendors have us scared to death about what we eat, drink and breathe. But most of their studies couldn't pass Statistics 101. • medical statistics By Philip E. Ross IN THE 1973 Ntovtr Sleeper, Woodv Allen plays a health food salesman who comes out of suspended animation in the year 2173 and calls for wheat germ, organic honey and tiger's milk. His puzzled doctor asks a colleague why people had once preferred such sludge to steak, cream pies, hot fudge and deep fat. "Those were thought to be unhealthy," the colleague replies, "precisely the opposite of what we now know to be true." Life imitates art. Twentv years too late, those of us weaned from real butter to oily margarine suddenly • learn that the syntheticalhl solidified oils of margarine, kno.vn as trans- fatty acids, were worse for our arter- ies than any fat found in nature. You would have been better off enjoying nature's spread, after all. Just because something tastes good doesn't mean it's bad for you. Such flip-flops in the history of health advice are the rule rather than the exception. Today's bad-for-you was probably once a good-for-you- and vice versa. Yet every ncw headline sends millions offon a search-and-destrov misslon in the pantrv or a panicky visit to the doctor. We have become a nation of nervous Nellies, ready to give up eating pleasures at the drop of a medical report. Mercury-laden fish is killing us, then coffee, then eggs, then it's too much chlorine in water, then it's too little. Even other people's pleasures are supposed to be danger- ous for us, like secondhand smoke. Strange, isn't it. We are living longer and healthier and are better nourished than ever before, yet imagine our- selves beset with poisons and cancer-causing substances. It's time for us to recognize that a whole industry exists to invent, propagate and then alleviate health scares. Medical schools get their grant money by publishing studies, the New England Journal of Medicine makes a living publishing them, popularizers make a nice living rewriting medical journal discoveries as books and articles. A large segment of the grocery products indus- trv makes it a business to sell so-called health foods. Ever notice all the intel- ligent-looking folks carefully scruti- nizing the fine print "nutritional" information on cereal boxes, jam jars and soup cans? We're not against folks making an honest dollar touting healthy diets, but as a publication with a healthy respect for statistics, we can't resist looking down our editorial nose at some of the statistical contortions hiding be- hind these scares. Let's start with the most amazing flip-flop of recent vears: alcohol. As the once aristocratic dry martini gave way to a glass of white wine, NIoung Americans came to drink much less than their parents and grandparents had. It had even become hip for some to teetotal. Then, last year, a group affiliated with the Harvard School of Public Health recommended that we consider taking a glass of wine a day. The libation was to be part of the Mediterra- nean Diet-so called because it mimics the low-meat, olive-oil-rich, wine-soaked regimen of southern Europe, where heart disease is relatively rare. Further supporting evidence came from laboratory findings that alcohol raises 130 Forbes . August 14, 1995

levels of high-density lipoprotein, the "good" cholesterol that rids the body of the artery-clogging kind. Pass the bottle? Not so fast. The trouble with most of these medical studies is that they look at a few factors in isolation. Okay, a little booze'is good for your heart. But what does it do for your chances of suffering cancer, stroke, cirrhosis of the liver or your chances of ending off the road in an overturned car? The moral of the story is: Drink if you like, but not because it will make you live longer. In fact, the New England Journal of Medicine-whose editorials continue to endorse the new drink-a-day-keeps- thc-doctor-away regimen-published a study in June showing that alcohol can raise blood pressure significantly. You can't make sense of such conflicting reports. Aren't medical journal articles peer-vetted, scientific and therefore unimpeachable? Don't believe it. A land- mark 1972 article in the journal Pediatrics almost single- handedly created an industry of sudden-infant-death pre- vention. A cluster of five cases in one upstate New York family supposedly showed that there was a genetic compo- nent to the syndrome; prevention included monitoring devices and varying advice for positioning sleeping babies in the crib. Whoops. Turns out that that New York mother mur- dered her five children. Pediatrics ran a retraction some 20 years later. By then moms and dads had lost a lot of sleep unnecessarily. Forl>es • August 14, 1995 131

Expect more such non- sense, because of the in- herent limitations in the epidemiology of nonin- fcctious disease. Scientists could devise an experi- ment that would really prove whether low doses of secondhand smoke hurt children, but the ex- periment would be uneth- ical. To really know whether exercise makes people live longer, experi- menters would have to randomly select 1,000 subjects and regiment their every waking mo- ment for decades. In default of such seri- ous studies, distrust most - ~ sweeping conclusions linking diet and health. Yes, you can prove statisti- cally that people who exercise live longer. But the conclu- sion-that exercise is good for you-may put the cart before the horse. Think about it. Are people healthy because they exercise? Or do they exercise because they are healthy? The bare correlation proves nothing. Physiologi- cal and other evidence leads many doctors to conclude exercise helps. But even here, there's no clear proof. The mass media purvey health scare stories breath- lessly, treating each new paper in isolation from the ones that came before. Even if the researchers concede that there are limitations to the validity of the study, the media usually bury the caveat. "You reporters take at face value everything the scientists tell you-something you would never think of doing if the source were industry or the government," grouses one prominent • epidemiologist. There's too much at stake for the health scare industry to admit to the shortcomings of epidemi- ology. The Berkeley, Har- vard, Johns Hopkins, Tufts and University of Texas medical schools are raising money by publish- ing health tip newsletters for the general public. The Journal of the Ameri- can Medical Association takes in $17 million in advertising a year. Jane Brody's Good Food Book: Living the High Carbohv- dratc Way (1985) is in its tenth printing. Sales of prepared foods making health claims ("low-fat," "no sugar added," "low sodium") account fbr $23 billion a vear. Let's go back to the butter/margarine argu- ment. In 1993 a group at Harvard linked trans-fattv acids to heart disease, adding, a year later, that the svnthetic fats mav be responsible for an extra 30,000 deaths a vear in the U.S. Margarine, of course, is full ofvcgetable oil hardened by a process called hydrogenation- which produces trans-fat- ty acids. The next vear, margarine sales dropped 8.2% by volume, and but- ter sales rose 1.4%. Take eggs. "You'd think there were halfa dozen studies saving to avoid eggs," says Walter C. Willett of the Harvard School of Public Health. "Guess how many there were? When we started looking at this question, zero. Now there's still very little information. No one's even shown it in an observational study [one where vou compare people who say they eat eggs with people who say they don't]. The recommenda- tion was based only on reasoning that cholesterol in diet affects blood cholesterol, and that eggs have cholesterol." Tell that to the chicken farmers. They've watched choles- terol-spooked consumers cut their egg consumption from a peak of 402 eggs per capita in 1945 to 238 today. The old eggs-are-bad theory was built on a pretty weak connection, that choles- terol in foods affects blood cholesterol. But the fact is that genetic make- up is far more important than consumption ofcho- lesterol in determining blood cholesterol levels. Meanwhile, millions of children who could bene- fit from the cheap protein in eggs are being shooed awav from it. Then there's the outra- geous campaign against "passive" smoking. Some studics have purported to show that people exposed to other people's smoke are more prone to lung cancer than the general population. The statistics are dubious. So-called passive smokers often 132 Forbes • August 14, 1995

~ So:what should I"do? _~C(~t~lb'ROtaTED with a bliz-- ~ard.:of conflicting advice aliout dietand health, :.:you, might either go crazy Mtrying to follow it all or -tht'ow•up your hands in ex- ; ~sp±ctation, pour yourself `,another beer and slump in =#`mrit of the w set with a. bowl of greasy potato chips. ~; No, don't do that, ci- ;'-t6.t;. Herc's.a middle - ;,10ound: Take •some of the advice seriously, Herewith .f~1 . _~ tii~efcompilation.of::,:.=-- Valth advice.according:to ~~hhthci it is well.proven; ~ ._.-. . aybe"Aen-ito .ry or-; • ably garbagc; ; : ; r'..: ~ ~S`~noking is bad foryou:: . Thid,ncds ~i:e .eyee „ , ~,:.. ;.:,. . 77 cnthiso ne, aio mat-,= hat~ihe-fobacco:com~:: statisticians say::: ; avy n~cing.is bad-: :: .. ±oq~ Eiioiigh alcohol-to. ' 1iu uPsif:con- °: r=. :ei~e,rji day; will grcat ' ; : e,.your_chahces of oais; cancer:and~n e=hi l gh y:p, eop• g.... : liTcx3d;picssurc:, . ~!'Sgn is,bad:for:you. A1- g~i'.you:needn't shield : ~o~iselfanlead foil every... you go. outside, sun . . ~ r-'. k:and a;Iiat~will.lower. k k:qfge[G[ng skin; ' Forbes . August 14, 1995 and thatprobablyelimi ,,; Here's what's probably ; mins C, E.and A~ave - natcs the statistical-.benefit':,,-r:°going on. Some people are ". produced mixed results.: ._' of light drinking. genetically sensitive to '  Overeating is bad I'or..•:' -" ; salt and would benefit from Bunk . you. Obese people definite- a low-salt diet, A lot,of '  There's no evidence '; ly don't live as:long as people aren't sensitive.and that eggs-are bad:for.yoii..: others. But no one has real=:: gain nothing. There's no You gain nothing bygiv= .:' ly proved that losing easy way to know which"ingxhem up in favor..of,oth-t weight will lengthen, hfe =R- group you fall into:` ' ~. -. er-forms ofsaturateda,. ' expectancy: ".. , Our advice: You willu-'•'.' Even researchers who en=,. `  Saturated fats seem to • probably do yourself no' :-: ; dorse the current wisdom contribute to hardening of harm avoiding salty: T;-i on.saturated fat say that an the arteries. If you are'at-;r,-e'snacks, and you might do:- °` . egg a day:'s not.bad::Ba-~: ;•. high risk for heart disease; some good. ;°- sides ahe.y contain high- - maybe you should cut •: w~ Fruits and vegetables -'. quality protein and back on these foods. Butif;'•,'.may lower, your chancc.of := they're cheap. ; M It's not wise to tat mar-i garine:_Ittums oufthat: butter-'.smaturallyy satarat-- : ed fat isn't as badas_-th4:. s; ; artifccial trans-fat'in most'' inargarines: =M Fish oil isn't thc hcart panacea it;was billcd.to be. :;:•The fish/heart connec-;; ,. tion was.debunked in re- . ~ .cent years:No wonder:..: One of'the•original study . .groups."was Greenlandic Eskimos; who differffi fro•cn ; us in a whole lott of ways ' ;,: besides diet.  There's no good evi-, dence that low-frequency > electromagnetic waves :, ... (from power lines and .; toasters), cause cancer. ;  There's no evidence .. that a• cup or two a day of coffee. is;bad for you. ~TJsrttltiiviolet-,t =- you're not a'lugh risk getting canccr::Itcould ' Agaimand again, attempts r~ fsun assesto ward off i.' case; ttiece'is na direct c%n be that vegetables contam ' to denyus this humble - ~t ~ y,c~afacts:~. '. dencc that such a' dierary4 antioxidants, which sop plcasure by means. of a sta- I~if~:`..'••'•-:'K change will improve yoiir;. '" up a chemically active form tistical survey havc been life span: of oxygen that seems to overturncd. ~ ittle alcohol a," N Salt may raise your cause cell damage: Or it 0 The EPn's conclusion a day=can lower blood pressurc: Circuni= could be that vegetables that 3,000 people a year die 1W:nsk ofcoronary ar-. -." stantial evidence from ' themselves don'.t do a from'secondhand.smoke sease: However, csoss-culturalstudies sug- . blessed thing; the con- ,., is disputed by reputable r it is wist too act '-` ':^':-gcsts that it may contrib', .. nection to cancer might :';,.':; epidemiologists. If '' :.. . . . . .. ~°eprrdaqon depends :. ute to hyperterision: Oathe': mcrely.be that vegetable. there's a Connection 6e-" ~ ;ethcr lighf diinking- -other hand, attempts'to '' eaters consuiiiie less fat-and ': tween aeconand' ~oii•r~;~skofsome -skofsome- _-_ .'r,;educe.blood,prsssure by fat contributes to cancer.,., ": -smoke and cancer,-it's an.' ~avy~*; s : mcans of ipw; saZtrdiets Recent studics of the mosr >'cztreinely :`pro`Yialily'does,.'" .: have bcen :'disappointing. likely antioxidants=vita P E.R ~ 133

share a lot of things besides their neighbors' snloke-like genetic heritage and social class ( FoitBrs, Jan. 31, 1994). Both are extremely relevant to cancer risk. In other words, the 1992 declaration bv the Environ- mental Protection Agency that secondhand snloke kills maybe 3,000 Americans a year is simply not credible. Some people even take seriously the fanciful connection between high-tension electric wires and "cancer clusters," a scare promoted by journalist Paul Brodeur in the Neii, Yorker in 1992 and in a book in 1993. Most epidenliolo- gists dismiss his scientific claims. "It's just garbage epidemiology," says Pro- fessor James Enstrom of UCLA. Are lefties condemned to short lives? In 1991 two psychologists, Stan- ley Coren of the Universi- • ty of British Columbia and Diane Halpern of California State Universi- ty in San Bernardino, questioned the relatives of deceased people in Cali- fornia and found that few- er and fewer left-handers showed up in the older age groups. They con- cluded that lefties were dying off at a greater rate. And they calculated that the average life expectan- cy of lefties was an as- tounding nine years less than that of righties. That's a bigger difference than that between smokers and nonsmokers. If true, it u•ould have easily been the greatest epidemiological relationship to have escaped notice up to that time. The Coren-Halpern findings ran in the New England Jozcrnal ofMedicine, giving them instant ~ credibility; the story even landed on the front page of the Washington Post. "I just laughed," says left-handed biostatistician Don- ald Hoover ofthe Johns Hopkins School of Public Health. "In fact, the older people had gone to school at a time when teachers forced lefties to write with their right hands. People of that generation weren't allowed to be l.ett- handed." Hence the "survev" had covered only a traction of those born left-handed. Furthermore, actuaries criti- cizcd the authors of the original study for mishandling their statistics. These subtleties escaped the journalists who rushed to publish the discovery of the supposed nine-year gap. Within a couple of years, professional journals such as the British Medical Journal and the American Journal of' PuGlic Hcaltb published articles debunking the Ictt\• hypothesis. But the Washinqton Post, so far as wc can tell, never did a fOlloW-up storv confess- ing that the earlier one was bunk. In the 1980s a number of papers came out show- lnf; a connection between cot}ee drinking and pan- creatic cancer. The link- age had no physiological theory to back it up, but it had scare value, consider- ing that tens of millions of Americans begin their day With cups of the stutt. C:otfee drinkers had about 2.5 times the risk of pan- creatic cancer as others, claimed one 1981 snidy. Its authors speculated that a bit more than half of all pancreatic cancers might stem from cotfee. The theori• made it into the Neu; England Journal (f :Llcdicinc. Relax and sip away. Later studies failed to reproduce the findings. Finally, in 1986, tthe authors of the original studv reported that further study had led tllenl to e:onclude that "if there is am, association between cot}ee ConSUnlptlOn and cancer of the pancreas, it is not as strong as our earlier data suggested." Translation: Coffee's as safe as orange juice, for all we know. How did this curious statistical correlation surtace? By a Berkeley;- thinks not•."`All ': ': 'y_: Do you jog) Have you _ Keep it up for'40 years; .~r _ this fuss about pcstici~e res-'~:figured out your expected idues is counterproduc-. : benefif in vears of life? ti.ve: If yoii drive aq extra .•: Now allow for your risk of milexogott;n"yourri'rgatiic:;_ gettiilg run over. It may . food store7ur nsk ofdy- ;~ fie-.a losing game. " ' ing in a _caracddcntis ; a: • jtreadmill is no better. crYthati~ : ;,~~gnsider the hour.a day, ose=-~ --Qfnree tunes a week, you.. - ' -~`^' -snend sweatina at the m~m. .s:., r.-. adding, say, a yea~ to.your lifespan. Guess:wliat :12°lie .: number of wakingFhours';r ~ .: vou uain in thatvearvt 'W:: life is less.than the nuriibei " of hours you' lostxo: ezer;', = cisc. Yes, exei c'iscr~sg43=; for you,:but_ovcrdoing isn't.. 134 Forbes m Augusr 14, 1995

• • scientific phenomenon known as data dredging, a creature of the age of huge databases and cheap computing power. Data dredgers can just take 100 food items and personal habits and com- pare them with 100 differ- ent ailments. If necessarv, divide and subdivide dis- eases like cancer accord- ing to the organ it affects. Now you have a matrix with 10,000 cells. Even if there is no cause-and-ef- tect relationship in amy of those cells, you are still sure to find around 100 stututing correlations about which you can hon- estlv make this claim: "The chance of this corre- lation arising by chance is less than 1 in 100." That's what you publish. Good science attempts to comprehend nature in a theory, then to make predictions on the basis ofthe theon,, then to test the predictions against observation or experi- ment. First you set up the target, then you shoot at it. Data dredging turns this on its ear. It finds data, then tries to build a theorv around it. It looks for bullet holes, draws circles around them and proclaims: bull's-eye. No, we're not saving that all correlations between diet and health are bunk. There's the connection between cigarette smoking and lung cancer. Statistics verified it across social classes and nations. Animal studies and tissue- culture experiments provided a physiologic basis for the connection. But too many people were quick to jump from this example to a broad indictment of every new substance introduced since the industrial revo- lution. Substances such as alar, formaldehyde and BHT are often alleged to lie behind the so-called ex- plosion in reported cancer cases. But, as manv epide- miologists have pointed out, there is no such epi- demic (except for lung cancer). Why then is can- cer increasing as industri- alization spreads? Because industrialization general- ly raises living standards, allowing more of us to live long enough to develop cancer and to allow a Forbes 0 August 14, 1995 slow-growing tumor to show itself: Industrializa- tion also invents and then pays for the high-tech tests that find cancers that would have eluded thr best diagnosticians of a generation ago. While survevs like the one going on in Franling- hanl, Mass. for the past-17 years can tell vou that high cholesterol correlates with the development of heart disease, that doesn't prove that lowering cho- lesterol will do anv good. A correlated variable nla\' be a marker, not a caUSe. Long hair correlates well with the female sex, but cutting a \voman's hair won't turn her into a illan. Even ifvou could deter- mine, once and filr all, that a given diet reduces vour risk of getting one disease, you still could not be sure that it would not raise vour risk of getting another. Sonlc doctors-a small minority, to be sure-believe that cho- lesterol-lowering programs push manv patients into dr- pression, perhaps for reasons of brain chemistn•. Ifthe\• arc right, sonle doctors nlav be driving people to suicide in the questionable name of reducing their risk of heart attack. A certain conservatism is in order here. Resist the temptation to incorporate a new finding into \'our hCalth regimen until it has been examined from nlanv angles, for nlanv vears. Wait out the long-term studies that actuallV test theories under controlled conditions. Othenwisc VOu nlav adopt a ctuc that's worse than the disease. To steel yourself in resisting such temptations, consid- er that hypochondria itselt is a disease, one that robs you not of life but Of its enjoyment. We're strange peoplc, we Americans. In 1900 our life expectancy .it birth was 47 vears. Now, thanks largely to im- proved sanitation, vac- ciiles and better 11VUriSh- meilt-all made \\'itjCll' available by our industri- al economv-our life ex- pectancy is 76 years. Do we rejoice at our good fortune? No, we are con- vinced a lot of people out there arc trying to poison us. ~ 135