in temperature was noted when he discontinued metoclopra-mide for 2 weeks. All investigations, including tests for sys-temic lupus erythematosus, rheumatoid arthritis, brucella andsalmonella infection, blood and urine cultures, and chestX-ray, were negative.When he stopped taking cimetidine his pyrexial episodes
became less frequent and ceased within 2 weeks. He has sinceremained well for 3 months, although his oesophagitis is caus-ing a little discomfort.
There is no evidence to suggest that this reaction was otherthan a transient drug-induced pyrexia.Department of Medicine,The Queens University of Belfast,Institute of Clinical Science,Belfast BT12 6BJ
J. C. MCLOUGHLINM. E. CALLENDERA. H. G. LOVE
SIR,-I am grateful to Dr Wynder (Feb. 4, p. 260) for indi-cating some of the limitations of classical epidemiology, es-pecially in the search for multiple causes of chronic conditionssuch as obesity, coronary heart-disease, and diabetes mellitus.The problem that confronts the non-specialist is how to inter-pret epidemiological studies. As an illustration, let me cite Bur-kitt. "The fibre hypothesis is frequently criticised becauseepidemiological associations do not prove cause and effect.This is true, but they do suggest hypotheses that can be tested.A lack of association between a disease and a postulated riskfactor does disprove the hypothesis." In other words, to thequestion-Is lack of dietary fibre a cause of obesity, C.H.D.,and diabetes?-a lack of association would answer "No" andthe existence of an association would answer "Maybe, so let uslook for other evidence".
Both of these answers need elaboration. A lack of associ-ation will not disprove the hypothesis of one suggested causewhen there are many. If we do not know what all the causesare, and if we cannot quantify the causes that we do know, itmay well be that we can find no association with one chosencause. No one, for example, has demonstrated a relationshipbetween the quantity or type of fat in the diet of individualsand their chance of acquiring C.H.D. ; the reason may be thatfat has nothing to do with causing the disease (as I believe),but it could be that we do not know enough about the relativeimportance of overweight, smoking, physical inactivity, andother known or probable factors, and that there are also un-known factors that of course cannot be measured.
So much for lack of association. What, then, of the presenceof association? To prove that an association is causal we mustdo experiments. Wynder says that animal studies are oflimited value because they are dimcult to extrapolate to man.Nevertheless this difficulty can be reduced by studying severalspecies of animals as well as man. The extension of classicalepidemiology to what Wynder calls metabolic epidemiologywould be much more useful if we could test by experimentwhether the metabolic disturbances discovered by epidemio-logy are produced by the suspected cause. In the same issue inwhich you published Wynders paper you included tworesearch reports (pp. 235, 238) and an editorial (p. 250) thatillustrate this.
Several years ago I suggested,2.3 largely from epidemiologi-cal considerations, that obesity, diabetes, and C.H.D. werecaused in part by the high consumption of sugar (sucrose) thatis characteristic of the diets of affluent countries. Since thattime my colleagues and I and several other research groupshave compared the effects of diets containing starch or sucrosein a number of species of animals and in man. Among theeffects of sucrose are the increased platelet adhesiveness and
1. Burkitt, D. P. Lancet, 1977, i, 1212.2. Yudkin, J. ibid. 1963, i, 1335.3. Yudkin, J. ibid. 1964, ii, 4.4. Szanto, S., Yudkin, J. Nature, 1970, 225, 467.
increased platelet aggregation in the presence of A.D.P., thatare reported, in your issue of Feb. 4, in diabetes mellitus.Other features of diabetes are also produced by dietary sucroseand include diminished glucose tolerance,6 hyperlipidaemia,tissue insensitivity to insulin,8 and the pathological changes inthe retina9 and kidney"" that characterise diabetes.The "refined carbohydrate" given as sucrose to the experi-
mental subjects was matched by the same quantity of refinedcarbohydrate given as pure starch or white bread to the con-trols. The effects of the sucrose have nothing to do with theabsence of dietary fibre but with the different ways in whichthe body reacts to starch and to sucrose.
Epidemiology is an invaluable tool in the search for thecause of disease, but it is increasingly necessary to add to it theresults of research. It is extraordinary that, despite the fre-quency with which workers insist that association is not proofof cause, so many of them proceed as if that is just what it is.
16 Holly Walk,London NW3 JOHN YUDKIN
LOW-DOSE HEPARIN AFTER STROKE
SIR,--Dr Buruma and his colleagues (Jan. 21, p. 160)report no hsemorrhagic complications from the low-dose sub-cutaneous heparin regimen they used in their stroke patients.However, their findings of no prophylactic effect on deep-veinthrombosis (D.V.T.) should be examined further.
Heparin acts by increasing the effect of antithrombin in,which is the major natural inhibitor of activated coagulationfactors. Low levels of heparin are sufficient to inhibit activatedfactor x, ix, and xi, enzymes which occur early in the coagula-tion sequence and can in small quantities lead to the formationof large amounts of thrombin. However, higher levels arerequired to prevent the formation of fibrin once thrombin hasbeen elaborated. Perhaps the heparin concentrations obtainedfrom low-dose subcutaneous heparin are insufficient to preventthe spread of thrombosis after the elaboration of the thrombinand consequent fibrin formation.The table summarising the results from Warlow et al., Den-
ham et al.1 and ourselves,3 shows the time at which the D.v.T.occurs, estimated by 1-fibrinogen scanning after acutestroke. The number of incidents of D. V. T. rose sharply after 48h and remained high for the first 10 days after the onset ofhemiplegia. If low-dose heparin is to be really effective itshould probably be given as soon as possible after the stroke,preferably within 48 h, and continued for at least 10 days.Buruma et al. do not state when their patients, other than thefour with thromboembolic complications, were started on ther-apy : one of their patients with complications first had heparinat day 8.Buruma et al. used a lower dose than we did. No large-scale
comparison of doses has been done. Measurement of heparinby the Denson assay shows that the concentration is maintainedabove 0.2 units/ml after 8 h but not at 12 h.4 An 8-hourlyregimen might be more effective because a therapeutic level ofheparin would be more likely to be maintained.
12(75%) of our 16 controls had positive leg scans. 7 weresuspected clinically, although only 3 had classical massiveoedema and pain. If the 25 Leyden patients were in the same
5. Bruckdorfer, K. R., Worcester, Nancy A., Yudkin, J. Nutr Metab. 1977, 21,suppl. 1, p. 196.
6. Cohen, A. M., Teitelbaum, A. Am. J. Physiol. 1964, 206, 105.7. Szanto, S., Yudkin, J. Postgrad. med. J. 1969, 45, 602.8. Bruckdorfer, K. R., Kang, S. S., Yudkij, J. Proc. Nutr. Soc. 1974, 33, 3A.9. Cohen, A. M., Michaelson, I. C., Yanko, L. Am. J. Ophthal. 1972, 73, 870.
10. Cohen, A. M., Rosenmann, E. Diabetologia, 1971, 7, 25.11. Kang, S. S., Price, R. G., Bruckdorfer, K. R., Worcester, Nancy A., Yudkin,
J. Biochem. Soc. Trans. 1977, 5, 235.1. Warlow, C., Ogston, D, Douglas, A. S. Lancet, 1972, i, 1305.2. Denham, M. J., James, G., Farran, K. Age Ageing, 1973, 2, 207.3. McCarthy, S. T., Turner, J. J., Robertson, D., Hawkey, C. J., Macey,
D. J. Lancet, 1977, ii, 800.4. Redman, C. Personal communication.