alcohol and cardiovascular disease

EDUCATION & DEBATE

Alcohol and cardiovascular disease: the status of the U shaped curve

Michael Marmot, Eric Brunner

Alcohol is responsible for considerable morbidityand mortality and for considerable social problems.Excess mortality associated with alcohol consumptionhas been estimated at 28 000 per year in England andWales,' and the annual cost of alcohol associated illnessin Britain was calculated to be 167 million in 1986;this is in addition to a social cost of about 1600million. The public health importance of limitingalcohol consumption is not questioned. This is thecontext within which to place the debate on theeffect of moderate levels of alcohol consumption onmorbidity and mortality associated with coronary heartdisease.

This debate has centred on the so called Ushaped curve describing the relation of alcohol con-sumption to risk of death from all causes. Heavydrinkers have an increased risk ofdeath compared withmoderate drinkers, but moderate drinkers have a lowermortality than abstainers. One commentator, sur-prisingly, labelled this U shaped curve a myth.' It is notmythical. Its existence is now not seriously in doubt. Itis due largely to the lower death rate from coronaryheart disease among moderate drinkers. What has beenquestioned is whether this is a consequence of aprotective effect of alcohol. Abstainers include peoplewho have stopped drinking because of ill health ("sickquitters") and people who are at higher risk for otherreasons. Either of these groups may account for thehigher risk among non-drinkers.The only major prospective study that failed to show

that moderate drinkers had a lower risk of cardio-vascular disease was that from Alameda County.' Onp 553 in a follow up to this study, Lazarus et al havereported that among the women former drinkers had ahigher risk of death than lifetime non-drinkers, butthis was not true for the men. Among the men lifetimeabstainers had a higher relative risk estimate thandrinkers, although the confidence limits overlapped.4A study from New Zealand reported recently inthis journal showed that moderate drinkers had alower incidence and mortality from coronary heartdisease compared with both former drinkers and neverdrinkers. 'What then is the status of the U shaped curve?

Does it represent causation? Is moderate drinkingprotective? Against the possible protective effect ofmoderate alcohol consumption on the risk of coronaryheart disease has to be set the direct effect on bloodpressure and the possible effect on risk of stroke.

Alcohol and coronary heart diseaseA study using data from 18 countries showed a

strong inverse association between average nationalwine consumption and mortality from coronary heartdisease.6 This type of ecological correlation, whichused a country's average value as the unit of analysis, issubject to the criticism that other factors may varyamong countries. The relation may not be specific toalcohol consumption. Interestingly, a more recent

international correlation study shows that the inverseassociation between alcohol intake and mortality fromcoronary heart disease is independent of the wellknown association of mortality from coronary heartdisease with fat intake. Indeed, adding mean alcoholintake to the multiple regression equation linkingfat intake to incidences of coronary heart diseaseimproved the correlation coefficient from 0 79 to 0 92.7These ecological data have been complemented bynumerous studies of individuals.

Several case-control studies, including that ofJackson et al, have shown that moderate drinkers ofalcohol have a lower risk of coronary heart diseasecompared with non-drinkers (fig 1). A review ofprospective studies in 1984 showed similarly thatmoderate drinking is associated with apparent protec-tion from coronary heart disease. 14 This has beenupdated to include studies published subsequently(fig 2). All of these studies, with, until now, theexception of the Alameda County study, found a lowerrate of coronary heart disease in moderate drinkersthan abstainers.The studies did not show uniformly that heavy

drinkers have a higher risk of cardiovascular diseasethan moderate drinkers. TheU shaped curve describingthe link of alcohol with risk of death comes from thehigher rate of death from various other causes in heavydrinkers.There is no clear evidence that one type of alcoholic

beverage is more protective than another, although theinternational study referred to above showed a strongercorrelation with wine.6

Interpreting the associationNATURE OF THE EVIDENCE

Each of the studies relating alcohol to cardiovasculardisease has its flaws. This is neither unique norsurprising. Despite the flaws of each individual investi-gation, what stands out in reviews of these studies'4 1 isthe consistency of the findings in populations as diverseas British civil servants, men in different regions of theUnited Kingdom, men and women in Framingham(Massachusetts), Yugoslavs, Puerto Ricans, JapaneseAmericans in Hawaii, Japanese physicians, WestAustralians, nurses in the north eastern United States,and others. Although consistency is often cited as oneof the criteria making an observed association morelikely to be causal than artefactual, studies can bewrong consistently. Nevertheless, where a largenumber of studies have used different methods butproduced similar results artefact is less likely; wherestudies have been performed in diverse populationsconfounding is less likely."SICK QUITTERS"One of the major criticisms levelled at many of the

studies is that non-drinkers will include people whohave given up drinking because they were unwell.'Such people would be expected to have an increased

BMJ VOLUME 303 7 SEPTEMBER 1991

Department ofEpidemiology and PublicHealth, University CollegeLondon, LondonWC1E 6EAMichael Marmot, FFPHM,professor ofepidemiology andpublic healthEric Brunner, MSC, researchfellow

BMJ 1991;303:565-8

565

rate of disease. The argument suggests that it is theinclusion of "sick quitters" with the non-drinkers thataccounts for the high incidence of coronary heartdisease in non-drinkers compared with moderatedrinkers.Many of the studies do not give data on lifetime

drinking patterns, but a few are able to separate themortality in lifetime abstainers. These are the studiesof people enrolled in the Kaiser-Permanente prepaid'health plan in California,32 which show never drinkersand former drinkers both to have higher incidences ofcoronary heart disease than moderate drinkers. Simi-larly, never drinkers among Japanese physicians2'and Japanese Americans20 have higher incidences ofcoronary heart disease than moderate drinkers. In theKaiser study former drinkers who stopped for medicalreasons had higher mortality from non-cardiovascularcauses but not cardiovascular causes.33 The report fromAlameda County in this issue looks specifically at theeffect of change in alcohol consumption on mortality.Among men lifetime abstainers had a higher relativerisk estimate than moderate drinkers, although theconfidence intervals overlapped. The study is some-

Men onlyAuckland5

Boston8

Cleveland, UK9

Kaiser Oakland10Men and Women

Boston collaborative studyl 1Seattle12

Women only

Auckland5

Los Angeles13

0 02 04 06 08 1.0 1 2Relative risk

FIG 1-Relative risk of coronary heart disease in moderate drinkers compared with non-drinkers in case-control studies. Bars show 95% confidence interval where available. (Moderate drinking= I drinking category rate ratio was calculated for highest consumption category)

Men onlyAlameda County4

Albanyl56IAmerican Cancer Society16

British regional heart study17 4Chicago West Electric18

Framingham19Honolulu20

Japanese physicians2lPuerto Rico sudden cardiac death22Puerto Rico myocardial infarction22Lipid Research Clinic, San Diego23 *

Trinidad24Whitehall 125Yugoslavia26

Men and womenBusselton, Australia27

Kaiser Oakland28 'AtL---Women only I

Alameda County4 *Boston nurses29Framingham199

Lipid Research Clinic, San Diego23 _ *0 05 1.0 1 5 20 25

Relative riskFIG 2-Relative risk of coronary heart disease in moderate drinkers compared with non-drinkers inprospective studies. Bars show 95% confidence interval where available. (Moderate drinking=

All causes - --1 7 Cancer -F---

Coronary heart disease *Accidents and violence -0 -

1.5 - Cerebrovascular diseases -a/-/1-313(I,

a)

0-7

None

7 Shaper A(. W\\alker M, annaiamethce (i. Alkohol and mortalitv in Britishimen: cxplinming theI -shapcd curve Lanitei 19868:ii:1267-73.

IS l)Dvc AR, Staiteilr J, IPatil (). Lecpper 7i1, Slhckclle RB, et al. Alcoholconsumption and 17-year mortality in the Chicago Westertt ElectricCompany stUtdy. 7 Pree lied 1980;9:78-90.

19 Friedman LA, Kimball AW. Coronarv heart dtiseasc mortalitv and alcohiilconsumpti(in in Framinghatm. Am 7 I7pidemilo 1986;124:481-9.

20 Y'ano K, Rholuds G(G, Kagan A. Collee, alcohol and risk of citronary hcartdiseasc anuiig Japanese iienii livitg in Hawaii. N Ingg7lj Mid 1977;297:405-9.

21 Koiti S, Ikeda M, 'Fokuidotnc S, Nishizuini M, Karatsune M1. Alcohol atidmortalitv: a cohort stUdy ot' male Japanese physicians. Int 7 l pidemiol1986;15:527-3 1.

22 Kittner SJ, Garcia-l'almieri MR. (C,ostas R, Cruz-Vidal M, Abbott RI), HaslikRJ. Alcoiiil aticl coroitars heart disease in Puerto Rico. Am '. fpidenilol1983;1 17:53X-51(.

23 Criqui MH, Cowan LD, Tyroler HA, Baiigdiwala S, Heiss G, Wallace RB,t ial. Lipioprotcinis as mecdiators for the effects of alcohol consumptioii atidcigarette stmokitig ott cardioxvascular mortality: rcsults from the lipidiescircliclicis foillii-up study. Am_7 Epidemiol 1987;126:629-37.

24 Mliller (,J. BeckIcs (GLA. .\laude (iH, Carson DC. Alcohol cotisutptitii:protectii)it agaitist coronary hcart disease and risks to hcalth. Intl7Ipide"miili1990;19:423-0.

25 MIarmot M6, Rose G, Sliplex .\lJ, 'l'homas BS. Alcohol and mortality: aui-sh.aped curse. Lan-et 1981 :i:S813-3.

h6 Ktozareii. I). \Votvodic N, (iordoii T, Kaelber C, McGee D, Zikcl %I'J.l)ritnkitig habits and death. Intj7 Ppidetniol 1983;12:145-50.

27 (ullen K, Stenlhouisc NS, \\earne KL. Alcohlol and mortality in the Busscltonstud\y. I,7hEpidemizol 1982;11:67-70.

28 Klatskv AL, Armstroiig MA, Fricdman Gi). Relationis of alciholic beverageuISC to stubsequcnt coronary arters discase hospitalizatioti. Am J Cardiol1986;58:7 10-4.

29 Stanmpfer MIJ, Colditz (A, Willctt WC, Speizer FE, Hcnnekens CH. Aprospective study of moderate alcohol conisumption and the risk of coronarydisease and stroke in women. N\Enzgl7 Med 1988;319:267-73.

30 Criqui MH. The redtuction of coronary heart disease sith light to moderatcalcohol consumption: eflfect or artifacti Brj A- ddict 1990;85:854-6.

31 Shaper AG. Alcohol and mortality: a review of prospectic stutdies. Br Adddict1990;85:837-47.

32 Klatsks AL, Armstrong MA, Friedinan GD. Alcohol and cardiovasculardeaths. Circulation 1989;80:61 1-4.

33 Klatsky AL, Armstrong MA, Friedman GD. Mortality in ex-drinkers.Circulation 1990;81:720.

34 Castelli WP, Doyle JT, Gordon T. Alcohol and blood lipids. The cooperativelipoprotein phenotyping study. Lancct 1977;ii: 153-5.

35 Mcadc TW', Chakrabarti R, Haities AP. Characteristics affecting fibrinolyticactivity and plasma fibrinogen concentrations. BRJ 1979;i: 153-6.

36 Dufls J(. Fallacy of the distributioni of alcohol consumption. Psychol Rep1982;50: 125-6.

37 Rose (G, Day S. The population mean prcdicts the number of dcviantindividuals. BMJfj 1990;301:1031-4.

I.sI ctcpied 29Juhs 1991)

Division of Social andBehavioral Medicine,University of Medicine andDentistry of New Jersey,New Jersey 08103, USDonaldW Light, PHD,pro]essorBMI 1991;303:568-70

Observations on the NHS reforms: an American perspective

Donald W Light

In the waiting rooms of family practices across the USpatients can pick up a prominently displayed, large,four colour, free Special Report on Health featuring astory of the Montagues, entitled, "Can you afford toget sick?"' Their daughter contracted leukaemia, and,although this middle class family has standard healthinsurance, the parents have spent the past three yearsdealing with scores of different claim forms, deduct-ibles, copayments, caps, exclusions, and erroneous(perhaps) denials in order to cope with the $500 000 ofbills for treating her.The report pictures Mrs Montague before a dining

table covered with seven piles of different corre-spondence over bills and claims; boxes full of files foreach doctor, laboratory, and hospital; and even aphotocopier she bought to help in the fight againstconstant threats of financial ruin. An accompanyingarticle features Dr Niles, a physician on the other sideof the private competitive insurance market.' Heis seen coping with myriad insurance forms andhundreds of provisos, protocols, or rules that insur-ance companies have elaborated as the counterparts tothe exclusions, deductibles, copayments, and caps forMrs Montague. Each elaboration serves to reducepayments by the insurers or to delay what has to bepaid so that the insurance companies can earn moreinterest on the premium income. All Mrs Montaguewants is to have the bills for her daughter Jenni paid.All Dr Niles wants is to be paid for the services herenders.With about 35 million Americans being uninsured

and about 70 million more suffering from private Swisscheese policies full of holes in coverage, at the sametime that medical expenses have surged past 12% ofthe gross national product and the $600bn mark, anAmerican observer is hardly in a position to sayanything critical about the NHS reforms. Yet it isperhaps just this experience with payer and providermarkets over the past 10 years (during which costs rosefaster than before the market reforms began) and aYankee bred outlook on human folly that may providea valuable perspective.

Upending an efficient systemTo start at the bottom line, the vigorous effort by

Mrs Thatcher, Mr Clarke, and officers in the Treasuryto wrench the NHS out of its allegedly wasteful,inefficient, and expensive ways seems strange to astudent of comparative health care systems. Do theynot know that they were (and are) getting better valuefor money than any other system in the industrialisedworld?3 No other system comes close in services perthousand pounds, even after taking into account themillion patients waiting for elective procedures.Do they not know that experts from other systems

marvel at how hard British nurses, technicians, andphysicians work for so little pay? The level of dedi-cation measurable in services rendered per millionpounds of pay provides this government with one ofthe greatest bargains in the world. Moreover, medicalexpenditures have been rising more slowly than in anycomparable system. Yet the government's responsehas been to purchase services at marginal rather thanaverage costs, forcing the whole NHS to become rundown, while all the time complaining about howexpensive the system is.

Ironically, this systemic parsimony has made theNHS steadily more inefficient: equipment breaksdown, inefficient or obsolete equipment is not replaced,arrangements are not replaced, poor working con-ditions and pay contribute to high staff turnover, and agrowing number of people opt out of the systemthrough private insurance whenever they can. Thusthe government is running down what many used toperceive as a competent, inexpensive, universal healthcare system. The alternative private system will bemuch more expensive. These private costs, however,will not show up on the bills that the government pays.To summarise, the restructuring of a system that

delivers more services per million pounds than anyother on the grounds that it does not do so seemsinsane. Were the cost of the NHS 45bn and rising at5% over inflation it would be understandable why thegovernment thought it had a problem; but with NHScosts at 30bn, rising at 2% over inflation, the questionis why do British leaders think they can treat thenation's ill for a third less than anyone else?

Although the governmnent apparently intends thereforms to make the NHS more efficient, their latentfunction is to transfer the political heat of parsimonyfrom politicians to managers and doctors in the guise of

568 BMJ VOLUME 303 7 SEPTEMBER 1991